PRIMES

MODEL
VERSION 2018
Detailed model description

E3Modelling
 PRIMES 2018

Contents
A. History ___________________________________________________________________________________ 1
B. General Overview _______________________________________________________________________ 2
C. Overview of Methodology ______________________________________________________________ 3
D. Stylized Mathematical Description ___________________________________________________ 11
D.1. Stylized Model ______________________________________________________________________ 11

D.2. Using PRIMES to meet policy targets: illustration _______________________________ 13

E. Policy focus _____________________________________________________________________________ 17

F. Typical Inputs and Outputs of PRIMES ______________________________________________ 20
G. Comparison to other models _________________________________________________________ 21
H. Overview of PRIMES model resolution ______________________________________________ 23
I. Data Sources and Model Linkages ____________________________________________________ 25
I.1. Data sources _________________________________________________________________________ 25

I.2. Model Linkages ______________________________________________________________________ 25

J. Stationary Energy Demand Sub-models _____________________________________________ 27

J.1. General Methodology _______________________________________________________________ 27

J.2. Industrial energy demand sub-models ____________________________________________ 32

J.3. Steam and heat generation _________________________________________________________ 43

J.4. PRIMES BuilMo: a detailed residential and services sector model ______________ 44

J.4.a. Model database _________________________________________________________________ 46

J.4.b. Equipment and electric appliances ___________________________________________ 48
J.4.c. Renovation strategies __________________________________________________________ 50
J.4.d. The heterogeneous consumer/actor __________________________________________ 51
J.4.e. Mathematical structure and model concept __________________________________ 52
J.4.f. Policy representation in the new buildings model ___________________________ 52
J.7. Mathematical Illustration of Modelling Energy Demand of Stationary Energy Use
Sectors in PRIMES model _______________________________________________________________ 57

J.7.a. First Level: Aggregate demand for energy service be sector ________________ 57
J.7.b. Formulation of allocation decisions of the consumer _______________________ 57
J.7.c. Second level: Uses or processes _______________________________________________ 59

Page i
 PRIMES 2018
J.7.d. Third level: Demand for technologies and fuels______________________________ 60
K. Transport Energy Demand (PRIMES-TREMOVE) ___________________________________ 63
K.1. Introduction ________________________________________________________________________ 63

K.2. Model overview ____________________________________________________________________ 63

K.3. Policy analysis focus of PRIMES-TREMOVE ______________________________________ 65

K.4. Novel Model Features ______________________________________________________________ 67

K.5. Demand and supply equilibrium in the transport model _______________________ 73

K.5.a. Overview _______________________________________________________________________ 73

K.5.b. The transport demand module _______________________________________________ 74
K.5.c. The transport services supply module _______________________________________ 78
K.5.d. Generalised Price of Transportation _________________________________________ 85
K.6. Refuelling/recharging Infrastructure _____________________________________________ 86

K.7. Calculation of external costs _______________________________________________________ 87

K.8. Measuring disutility costs _________________________________________________________ 87

K.9. Source of data and calibration to statistics _______________________________________ 88

K.10. Classification of transport means ________________________________________________ 90

K.11. Model outputs _____________________________________________________________________ 93

K.12. Transport activity modelling using econometrics in v. 6 ______________________ 94

K.12.a. Transport activity projections_______________________________________________ 94

K.12.b. Data update and calibration of the transport sector ______________________ 95
K.13. PRIMES-Maritime transport model _____________________________________________ 95

K.13.a. Introduction __________________________________________________________________ 95

K.13.b. Processing of data input _____________________________________________________ 98
K.13.c. Supply- Fleet module ________________________________________________________ 101
L. Power and Steam Generation and Supply Models_________________________________ 105
L.1. Overview ___________________________________________________________________________ 105

L.2. Mathematical Structure ___________________________________________________________ 107

L.3. Model Features ____________________________________________________________________ 111

L.3.a. Representation of Plants _____________________________________________________ 111

L.3.b. Investment modelling ________________________________________________________ 111

Page ii
 PRIMES 2018
L.3.c. Blending of fuels_______________________________________________________________ 111
L.3.d. Scale of generation activity___________________________________________________ 112
L.3.e. CHP _____________________________________________________________________________ 112
L.3.f. CCS ______________________________________________________________________________ 112
L.3.g. Nonlinear cost curves _________________________________________________________ 113
L.3.h. Plant dispatching and system operation ____________________________________ 113
L.3.i. Environmental Policies________________________________________________________ 114
L.4. Data sources of PRIMES power and steam model_______________________________ 114

L.5. Modelling of the EU power network and market _______________________________ 115

L.6. List of plant technologies _________________________________________________________ 116

L.7. Technology Progress ______________________________________________________________ 118

L.8. Investment and Renewables ______________________________________________________ 118

L.9. Investment and Nuclear Energy __________________________________________________ 121

L.10. Investment and CCS ______________________________________________________________ 122

L.11. Financial and Pricing Model for Electricity, heat and steam __________________ 123

L.13. Simulation of Oligopoly Competition in Electricity Markets __________________ 125

L.14. Power sector sub-model of PRIMES Version 6_________________________________ 127

L.15. Special version of power sector model for the Internal European Market: The
PRIMES/IEM model ____________________________________________________________________ 129

L.15.a. Introduction to PRIMES/IEM model _______________________________________ 129

L.15.b. Modelling procedure ________________________________________________________ 130
L.15.c. Day-ahead market simulator (DAM_Simul)________________________________ 131
L.15.d. Network representation in DAM ___________________________________________ 132
L.15.e. Bidding of power plants _____________________________________________________ 132
L.15.f. Modelling of nominations ___________________________________________________ 132
L.15.g. Modelling of priority dispatch ______________________________________________ 133
L.15.h. Modelling of demand response _____________________________________________ 133
L.15.i. Day-ahead market simulator version with Unit Commitment (DAUC) __ 133
L.15.j. Random Events Generator (optional) ______________________________________ 133
L.15.k. Unit Commitment simulator (UC_Simul) __________________________________ 134
L.15.l. Intra-Day and Balancing simulator (IDB_Simul) ___________________________ 135

Page iii
 PRIMES 2018
L.15.m. Reserve and Ancillary Services market or procurement simulator
(RAS_Simul) __________________________________________________________________________ 137
L.15.n. Mathematical Illustration of PRIMES-IEM model _________________________ 138
L.15.o. Methodology of the Modelling of Power Generation Investment under
Uncertainty (optional) ______________________________________________________________ 141
M. The PRIMES Gas Supply Model __________________________________________________ 145
M.1. Scope of the model ________________________________________________________________ 145

M.2. Gas infrastructure_________________________________________________________________ 145

M.3. Modelling of competition _________________________________________________________ 146

M.4. Model usage _______________________________________________________________________ 148

N. Oil Products Supply Model and biofuel blending __________________________________ 150

O. Rest of Energy Branch related to fossil fuels ______________________________________ 152
P. Projection of Energy Balances ______________________________________________________ 153
Q. The PRIMES Biomass model ________________________________________________________ 157
Q.1. Model scope and aim ______________________________________________________________ 157

Q.2. Structure, feedstock and conversion technologies _____________________________ 157

Q.2.a. Structure ______________________________________________________________________ 157

Q.2.b. Feedstock______________________________________________________________________ 158
Q.2.c. Biomass Conversion __________________________________________________________ 159
Q.2.d. Technologies for Bioethanol production ___________________________________ 160
Q.2.e. Technologies for Biodiesel production______________________________________ 161
Q.2.f. Technologies for bio-kerosene production__________________________________ 161
Q.2.g. Technologies for biogas and bio-methane production_____________________ 162
Q.2.h. Technologies for bio-heavy production _____________________________________ 162
Q.2.i. Technologies for Small & Large Scale production of solids ________________ 162
Q.2.j. Technologies for bio-hydrogen production _________________________________ 162
Q.3. Biomass Model Methodology _____________________________________________________ 163

Q.3.a. Inputs __________________________________________________________________________ 163

Q.3.b. Endogenous trade ____________________________________________________________ 164
Q.3.c. Mathematical specification of the model ____________________________________ 164
Q.4. Representation of policies and measures _______________________________________ 168

Q.5. Database of PRIMES biomass model _____________________________________________ 169

Page iv
 PRIMES 2018
Q.6. Biomass Conversion Chains ______________________________________________________ 175

R. PRIMES New Fuels Model ___________________________________________________________ 190

S. Greenhouse Gas Emissions and Policies ___________________________________________ 194
S.1. Emissions ___________________________________________________________________________ 194

S.2. Emission Reduction _______________________________________________________________ 194

S.3. ETS Market Simulation ____________________________________________________________ 195

T. Prices of Energy Commodities ______________________________________________________ 198

T.1. Introduction _______________________________________________________________________ 198

T.2. Mathematical illustration of price calculation___________________________________ 199

U. PRIMES reporting on Energy System Costs ________________________________________ 201

V. Methodology on Discount Rates ____________________________________________________ 202
V.1. Overview of discount rates within a modelling approach ______________________ 202

V.2. Capital budgeting decisions in PRIMES __________________________________________ 203

V.3. Methodology for defining values of discount rates _____________________________ 204

V.3.a. Decisions by firms generally follow the approach of the weighted average
cost of capital (WACC) to define discount rates. __________________________________ 204
V.3.b. Decisions by individuals using a subjective discount rate to annualize
investment (upfront) costs following the equivalent annuity cost method. ____ 205
V.3.c. Discount factors used to evaluate tariffs of using infrastructure regulated as
a natural monopoly. _________________________________________________________________ 208
V.3.d. Business sectors ______________________________________________________________ 208
V.3.e. Households ____________________________________________________________________ 209
V.4. Use of discount factors for energy system costs reporting _____________________ 209

V.4.a. Overview ______________________________________________________________________ 209

V.4.b. Present Value Calculation Method __________________________________________ 211
V.4.c. Mathematical illustration of cost reporting in PRIMES ____________________ 213

Page v
 PRIMES 2018

A.History
The PRIMES (Price-Induced Market Equilibrium System) energy
system model is a development of the Energy-Economy-
Environment Modelling Laboratory at National Technical University
of Athens in the context of a series of research programmes co-
financed by the European Commission. The model has been
successfully peer reviewed in the framework of the European
Commission in 1997 and in 2012. The techno-economic
parameters of the PRIMES model were recently reviewed by a
broad range of stakeholders within an ASSET project study.1
From the very beginning, in 1993-1994, the design of the PRIMES
energy model focused on market mechanisms and aimed at
explicitly projecting prices, which influence the evolution of energy
demand and supply as well as technology progress. The model
structure is modular. The modules differ by sector in an aim to
represent agent behaviours and their interactions within the
markets as close as possible to reality. The model design
combines microeconomic foundation of behaviours with
engineering and technology details. The mathematical
specification focuses on simulation of structural changes and long-
term system transitions, rather than short term forecasting.
From mid-90s until today, the model is regularly extended and
updated. Numerous studies have been performed using PRIMES,
and numerous third party studies have used projections produced
using PRIMES. The majority of these studies focused on medium
and long term restructuring of the EU energy system, aiming at
reducing carbon emissions. PRIMES supported analysis for major
energy policy and market issues, including electricity market, gas
supply, renewable energy development, energy efficiency in
demand sectors and numerous technology specific analysis, such
as on CCS, nuclear, etc. The PRIMES model has quantified energy
outlook scenarios for the EU (Trends publications since 1990), the
latest being the “Reference scenario 2016”, impact assessment
studies for the EC, including for the Clean Energy Package for all
Europeans, as well as the work for the Mid-century Strategy
(forthcoming end 2018). PRIMES also supported national
projections for governments, companies and other institutions
including for EURELECTRIC, EUROGAS and many others.

1 https://ec.europa.eu/energy/en/studies/review-technology-assumptions-decarbonisation-scenarios

Page 1
 PRIMES 2018

B.General Overview
PRIMES provides detailed projections of energy demand, supply, prices and
General aims of
investment to the future, covering the entire energy system including
PRIMES
emissions for each individual European country and for Europe-wide trade of
modelling: energy commodities.
 Market The distinctive feature of PRIMES is the combination of behavioural modelling
orientation following a micro-economic foundation with engineering and system aspects,
covering all sectors and markets at a high level of detail.
 Focus on
behaviors PRIMES focuses on prices as a means of balancing demand and supply
simultaneously in several markets for energy and emissions. The model
 Economics and determines market equilibrium volumes by finding the prices of each energy
Engineering form such that the quantity producers find best to supply matches the quantity
combined consumers wish to use.

Investment is generally endogenous in PRIMES and in all sectors, including for

 Long-term
purchasing of equipment and vehicles in demand sectors and for building
restructuring of
energy producing plants in supply sectors. The model handles dynamics under
energy systems
different anticipation assumptions and projects over a long-term horizon
 Explicit price keeping track of technology vintages in all sectors. Technology learning and
economies of scale are fully included and are generally endogenous depending
and costs
on market development.
projections
PRIMES model design is suitable for medium- and long-term energy system
 Full coverage of projections and system restructuring up to 2070, in both demand and supply
all energy sides. The model can support impact assessment of specific energy and
sectors and environment policies and measures, applied at Member State or EU level,
markets, including price signals, such as taxation, subsidies, ETS, technology promoting
including ETS policies, RES supporting policies, efficiency promoting policies, environmental
policies and technology standards. PRIMES is sufficiently detailed to represent
 Individual concrete policy measures in various sectors, including market design options
modelling of all for the EU internal electricity and gas markets. Policy analysis draws on
European comparing results of scenarios against a reference projection.
countries and
EU trade The linked models PRIMES, GEM-E3 and IIASA’s GAINS (for non-CO2 gases and
air quality) perform energy-economy-environment policy analysis in a closed-
loop.

Page 2
 PRIMES 2018

C. Overview of Methodology
The PRIMES model comprises several sub-models (modules), each one
Basic representing the behaviour of a specific (or representative) agent, a demander
methodological and/or a supplier of energy. The sub-models link with each other through a
ideas: model integration algorithm, which determines equilibrium prices in multiple
markets and equilibrium volumes meets balancing and overall (e.g. emission)
 Modular constraints.
organization
Mathematically PRIMES solves an EPEC problem (equilibrium problem with
 EPEC with equilibrium constraints) which allows prices to be explicitly determined.
explicit prices
The agents’ behaviours are sector-specific. The modelling draws on structural
 Micro-economic microeconomics: each demand module formulates a representative agent who
maximises benefits (profit, utility, etc.) from energy demand and non-energy
foundation
inputs (commodities, production factors) subject to prices, budget and other
 Concurrent constraints. The constraints relate to activity, comfort, equipment, technology,
energy and non- environment or fuel availability. The supply modules formulate stylised
energy inputs companies aiming at minimising costs (or maximising profits in model variants
focusing on market competition) to meet demand subject to constraints
 Hybrid model, related to capacities, fuel availability, environment, system reliability, etc.
embedding
PRIMES is a hybrid model in the sense that it captures technology and
technologies in engineering detail together with micro and macro interactions and dynamics.
economic Because PRIMES follows a structural modelling approach, in contrast with
decisions reduced-form modelling, it integrates technology/engineering details and
constraints in economic modelling of behaviours. Microeconomic foundation is
a distinguishing feature of the PRIMES model and applies to all sectors. The
modelling of decisions draw on economics, but the constraints and possibilities
reflect engineering feasibility and restrictions.

The model thus combines economics with engineering, ensuring consistency in

terms of engineering feasibility, being transparent in terms of system
operation and being able to capture features of individual technologies and
policies influencing their development. Nevertheless, PRIMES is more
aggregated than engineering models, but far more disaggregated than
econometric (or reduced form) models.

The model performs analytical cost estimations and projections by sector both
in demand and supply, as well as for infrastructure. Supply-side modules
determine commodity and infrastructure prices by end-use sector (tariffs) by

Page 3
 PRIMES 2018
applying various methodologies by sector as appropriate for recovering costs
depending on market conditions and regulation where applicable.
Prices and
Equilibrium: Prices influence demand and demand influences in turn supply. Thus, a closed-
loop between demand and supply solves simultaneously for all markets. Both
 Tariffs and demand and supply modules may be subject to system-wide constraints,
prices are mirroring overall targets for example on emissions, renewables, efficiency,
endogenous, import dependency, etc. The demand and supply modules are subject to
reflecting costs system-wide constraints, which when binding convey non-zero shadow prices
and market (dual values) to the demand and supply modules. Therefore, the PRIMES
conditions model has overall a mixed-complementarity mathematical structure. The
overall convergence algorithm simultaneously determines multi-market
 Closed-loop equilibrium while meeting the system-wide constraints.
between
The agents are a priori price-takers when being energy demanders and price-
demand and
makers when being energy suppliers. Optionally the model can handle non-
supply
perfect market competition regimes. The electricity and gas market modules
optionally include explicit companies (or stylised companies) and apply Nash-
 System-wide
Cournot competition with conjectural variations.
constraints
influence all In the demand sub-models, the agents are simultaneously self-producers of
sectorial sub- energy services (e.g. using a private car, heating using a residential boiler, etc.)
models and purchasers of marketed energy commodities. The pricing of self-supplied
energy services is endogenous and reflect average total costs. The mix of self-
 Self-supply of supply and the purchasing from external suppliers (e.g. private cars versus
energy services public transportation, residential boiler versus district heating) derives from
is also priced agent’s optimisation.

 Perceived costs Pricing and costing include taxes, subsidies, levies and charges, congestion
and uncertainty fees, tariffs for use of infrastructure etc. Usually these instruments are
factors are exogenous to the model and reflect policy assumptions. The model handles
included and are endogenously cap and trade policies and policies reflecting obligations. The
cap and trade policies (for example trading of emission allowances, green
related to
certificates and white certificates) involve issuance of certificates (or permits)
policies
and trading rules. The model projects certificate prices of equilibrium as result
of simultaneous equilibrium of all markets. The model represent obligations,
such as renewables or energy efficiency targets, as constraints. The model
estimates the shadow prices associated to such constraints, and includes them
in demand and supply sub-models where appropriate.

Some cost components are subjective reflecting uncertainty and perception

about performance and cost of advanced, not yet mature, technologies.

Page 4
 PRIMES 2018
PRIMES follows a descriptive approach concerning factors which influence
decisions by private entities where perceived costs and uncertainty factors
play a significant role. Policy measures can reduce uncertainty and decrease
Dynamics and perceived costs: such a mechanism in the model is often used to simulate
technology: policy inducing higher uptake of advanced technology or investment enabling
accelerated energy efficiency progress.
 Dynamic model
The PRIMES model is fully dynamic and has options regarding future
 Usually perfect anticipation by agents in decision-making. Usually, PRIMES assumes perfect
foresight, but foresight over a short time horizon for demand sectors and perfect foresight
other options over long time horizon for supply sectors. The sub-models solve over the
are available entire projection period in each cycle of interaction between demand and
supply and so market equilibrium is dynamic and not static. Other options are
 Investment and available allowing the model user to specify shorter time horizons for
equipment foresight.
choice is
All economic decisions of the agents are dynamic and concern both operation
endogenous in
of existing equipment and investment in new equipment, both when
all sectors
equipment is using energy and when it is producing energy.
 Technology Capital formation derives from economically driven investment and follows a
vintages are dynamic accounting of equipment technology vintages: equipment invested on
traced a specific date inherit the technical-economic characteristics of the technology
vintage corresponding to that date. Capital turnover is dynamic and the model
 Technology keeps track of capital vintages and their specific technical characteristics. The
progress agent’s investment behaviour consists in building or purchasing new energy
develops equipment to cover new needs, or retrofitting existing equipment or even for
through the replacing prematurely old equipment for economic reasons.
vintages and is
All formulations of agent behaviours consider technologies, which are either
endogenous
existing at present or expected to become available in the future. The
 Infrastructure technology selection decisions depend on technical-economic characteristics
influences and of these technologies, which change over time either autonomously
constraints (exogenous) or because of the technology-selection decisions (learning and
decisions but is scale effects). Perceived costs associated to technologies may change in
synchronised manner with technology uptake and learning.
exogenous to the
model The outcomes of decisions by sector generally depend on availability of
infrastructure and the usage tariffs. The model projects infrastructure tariffs
for cost recovering using regulated discount rates. Availability of
infrastructure influences technology uptake where applicable.

Page 5
 PRIMES 2018
Investment in network infrastructure is exogenous to the model but it can vary
by scenario. It includes electricity grids, smart systems, gas infrastructure, CO2
Capital budgeting transportation and storage, refuelling and recharging infrastructure in
decisions: transport sector.

 Capital The agents’ investment for energy production, the purchasing of durable goods
budgeting by consumers and the energy saving expenditures in buildings and houses are
decisions at simulated as capital-budgeting decisions for new investment, possible
various levels premature scrapping of old equipment or for retrofitting old equipment.
Retrofitting depends on specific costs and scrapping depends on maintenance
 Capital stock and variable costs, which increase over time because of ageing. Investment
accounting and scrapping decisions are included in accounting for the dynamics of
depends on capacity stocks in all sub-models.
investment,
The capital budgeting decisions refer to choices with different distributions of
possible
fixed and variable costs over time. The choices depend on annuity payments
premature
for investment expenditures, which in turn depend on interest rates, which are
scrapping and specific to each agent (sector).
eventual
retrofitting PRIMES follows a descriptive approach to the modelling of interest rates based
on the opportunity cost of drawing funds from individuals or private
 Discount companies. Interest rates are calculated based on the concept of WACC
(interest) rates (weighted average cost of capital), which involve a basic risk-free interest rate
based on WACC applied on equity capital, a bank lending interest rate applied on the part of
reflecting capital borrowed and a risk premium. All rates are net of inflation.
private
The interest rates applied on equity capital reflect agent-specific subjective
investor’s rates and are sector-specific. Risk premiums apply with two components: one
perspective specific to each sector and one specific to the candidate technology. For the
latter the model considers that innovative technologies that may not be
 Technology and
sufficiently mature or that may not dispose a sufficiently broad maintenance
market related
service support are more risky than market-established technologies.
risk premium
Different scenarios quantified using PRIMES may imply different distributions
 Social discount of costs over time. To compare them and to aggregate system-wide costs over
rates apply on time a present value method applies as a calculation external to PRIMES. The
calculation of comparison of performance across scenarios uses aggregation of costs over
present value of time, which by default uses a social discount rate. This rate differs in nature
system-wide from the interest rates used by sector to annualise investment expenditures
costs over time and to compare choices from a private investor perspective. The sector-
specific interest rates reflect opportunity costs of raising funds by private
entities and the social discount rate reflects opportunity costs of raising funds

Page 6
 PRIMES 2018
by the public sector. The discount rates are exogenous and can vary by
scenario.
Energy Demand
sectors: For each sector, representative agents optimise an economic objective
function: utility maximisation for households and passenger transport and cost
 Demand for minimisation for industrial, tertiary and freight transport sectors.
energy derived
To optimise, the model firstly considers useful energy demand (end-use
from
energy services) and then a nesting of further decisions. At the upper level of
microeconomic
the nesting, energy is a production factor or a utility providing factor and
optimization
competes with non-energy inputs. Useful energy, as derived at upper level,
decomposes into uses and processes (e.g. water heating, motor drives,
 Detailed and
industrial processes, etc.). Useful energy (e.g. air conditioning, lighting, motive
transparent
power) fulfils by consuming final energy, which derives from optimisation
demand
involving self-supply, purchasing of marketed commodities and investment in
formation from
equipment. Each demand model involves an internal demand and supply loop
end-use services
formulated in mixed complementarity mathematical structure. The self-supply
up to final is dynamic over time involving endogenous choice of equipment (vintages,
energy and use technologies and learning), endogenous investment in energy efficiency
of equipment (savings), endogenous purchase of associated energy carriers and fuels
(demander is price taker). Mathematics based on discrete choice theory
 Endogenous
captures heterogeneity within each representative agent. Decisions at each
choice of
nesting level uses relative costs based on equivalent perceived cost, reflecting
technology and actual costs, utility (e.g. comfort) and risk premium.
energy savings
investment Industrial energy demand modelling starts from projecting physical output;
the model focuses on materials, process flow and efficiency potential. The
 Focus on process flows include a variety of stylised industrial processes. The model
materials, distinguishes between scrap/recycling processes and basic processing for iron
efficiency and and steel, aluminium, copper, glass and cement (own production of clinker
processing in versus import of clinker). The process flows recycle industrial by-products
industrial sub- such as black liquor, blast furnace gas, etc. Energy saving possibilities depend
models on capital turnover, which his dynamic and endogenous. The possibilities are
specific to the current and future technologies, which are available for each
 Focus on type of industrial process. The model includes possibility of shifts towards
efficiency and more efficient process technologies and horizontal processing measures.
renovation of Interaction with Power and Steam sub-model for industrial CHP and boilers
buildings/house performs through the model-integrating module. Substitutions are possible
s and on between processes, energy forms, technologies and energy savings.
appliances For residential and tertiary sectors, multiple substitutions are possible. Useful
energy demand depends on behavioural characteristics partly influenced by

Page 7
 PRIMES 2018
costs and prices. The model includes a distinction of households types
according to energy consumption, it further distinguishes agriculture and
Supply system services which are broken down by sub-sector (e.g. market services, trade);
operation and electric appliances are treated separately in all sectors. Final energy demand
networks: depends on thermal integrity of buildings, with consideration of renovation
investment (several categories) and vintages. The model includes heat pumps
 Simulation of
and direct use of RES.
system
operation in Demand-related decisions at all levels depend on a large variety of policies,
supply-side which are explicitly represented.

PRIMES is very detailed in energy supply sub-models aiming at representing

 Demand varies
system operation aspects, related to interoperability between production units
over time and
and transportation infrastructure over networks and other means including
season for
storage, and reliable delivery of energy to time-varying demand when storage
electricity, gas
is limited, such as for electricity, gas and distributed heat/steam.
and distributed
heat/steam Load curves (chronological covering typical time-periods by year) for each
carrier (electricity, gas, distributed heat/steam) with time-varying demand
 Intermittent RES derive in the model in a bottom up manner depending on the load profiles of
are simulated individual end-uses of energy. Smart metering and other load and demand
management measures are included aiming at influencing demand variability.
 Trade in the
European In the simulation of electricity system operation PRIMES takes into account the
internal market intermittency features of renewable sources. Although it represents renewable
is endogenous sources in a deterministic manner, the model captures balancing, flexibility
and reserve-power requirements.
 System PRIMES models in detail trade of electricity and gas across countries deriving
operation from simulation of Europe-wide interconnected systems including full
depends on modelling details by individual country.
network
infrastructure PRIMES synchronises demand variability between electricity and distributed
heat/steam and captures operation of cogeneration units, which produce both
 Cogeneration, electricity and heat depending on both markets.
industrial units
The simulation of power and heat/steam markets includes competition
and highly
between plants for different purposes (pure electric, CHP, industrial boilers),
distributed takes into account networks in power and steam/heat markets and represents
production are plant economics by scale and aim, distinguishing between utilities, industries
represented and highly distributed generation. Self-production (by industries or
simultaneously individuals) is an endogenous possibility among the options and has distinct
with utility plant economics and dependency on grids.
supply

Page 8
 PRIMES 2018
PRIMES represents competing storage technologies and simulates their
operation in the supply systems. Investment in storage is endogenous driven
by economics.

PRIMES includes all other fuel supply sectors, including extraction, imports,
briquetting, liquefaction/gasification, bio-energy conversion, synthetic gas,
hydrogen and refineries. PRIMES generally involves non-linear formulations:

 Useful energy demand involves saturation levels and uses non-linear

formulas
 The energy demand models formulate nested budgeting and involve
non-linear indifference and isocost curves
 Models of discrete technology choices are non-linear (e.g. using Weibull
or logit functions)
 Economics of scale and learning-by-doing are non-linear by nature
 Costs related to potentials of resources (e.g. renewables) or to
possibilities of energy savings (e.g. energy efficiency measures) are
represented as non-linear cost-quantity functions
 In power system optimisation non-linear cost-curves represent fuel
supply, renewable potentials and limitations on development of new
power plant sites, where applicable (e.g. nuclear plant sites, wind sites,
etc.)
 Similarly storage potential including for CO2 storage involve non-linear
cost curves
 Cost of infrastructure depends on features such as integration of RES,
high distribution etc. in a non-linear manner.

PRIMES Modelling Scheme

Demand = function of Price

Through fairly complex energy demand projection models by sector
Supply = Demand
Through complex energy supply models with system operation and
network details
Price = function of Supply
Through a finance and pricing model which reflects costs, market
competition regime and regulation
System-wide targets
They influence all sub-models which see shadow prices associated to
targets
Iteration on Prices and shadow prices until reaching equilibrium
Iterations follow a Gauss-Seidel algorithm

Page 9
 PRIMES 2018

Summary of
PRIMES modelling Exogenous
approach  Economic Activity
 World energy prices
 Technology parameters
 Policies and measures

Sequence of model interactions

 Agents (representative household, industry per sector, services, power
generation, etc.) act individually optimizing their profit or welfare, influenced
by habits, comfort, risk, technology, system reliability, etc. using individual
(private) discount rates for capital-budgeting choices
 Accordingly they determine energy flows, investment and choice of explicit
technologies in vintages
 Demand and supply of energy commodities interact with each other over a
market with assumed competition regime
 Simultaneous energy (and emissions or certificate) markets are cleared to
determine prices that balance demand and supply
 Commodity tariffs reflect costs and apply a Ramsey-Boiteux methodology to
recover fixed costs and determine a distribution of tariffs across sectors
 Market equilibrium spans over the entire time horizon with investment being
endogenous
 Overall or sectorial restrictions apply, for example on carbon dioxide
emissions or for other targets

Mathematically, the model solves as a concatenation of mixed-complementarity

problems with equilibrium conditions and overall constraints (e.g. carbon constraint
with associated shadow carbon value); this is an EPEC problem.

Foresight is built in the agents’ decision-making representations, depending on

lifetime of equipment with market equilibrium being intertemporal.

Explicit technologies are included in all demand and supply sectors

 Technology dynamics
 Vintages
 Penetration of new technologies
 Inertia from past capital stocks and in future from capital turnover

Page 10
 PRIMES 2018

D. Stylized Mathematical Description

 Consumers get utility using energy and non-energy goods and services, including energy
Summary of efficiency as a means of meeting useful energy demand.
PRIMES modelling  Producers of energy carriers such as electricity, gas, distributed heat or hydrogen mix
approach through optimization fossil fuels and clean energy forms (e.g. renewables, nuclear or
carbon capture and storage) to produce the amounts demanded by consumers.
 They set prices of energy carriers to reflect total production costs. Consumers are price
takers but price-elastic.
 The primary energy sources, which are the fossil fuels, the clean energy forms used by
consumers and those used by energy carrier producers, use prices depending on cost-
supply curves with a positive slope and exhaustible potential.
 The consumers of primary energy forms are assumed price takers.
 Demand and supply behaviours are balanced in simultaneously clearing markets at
primary, carrier and final energy levels
 Overall and sectoral policy constraints may apply, e.g. regulations, emission targets,
renewable targets, energy efficiency targets
 The consumers and producers see the constraints through their associated shadow values
(e.g. marginal costs) which are found different from zero if they are binding at equilibrium
 All decisions also involve capital budgeting choices, hence determine investment using
technologies with features changing over time (vintages); capital stock is dynamically
updated exerting capacity constraints on flows
 Thus, decisions depend on anticipation about future evolution; the model applies perfect
or partial foresight.

D.1. Stylized Model

Consumers (problem (1)) maximize utility (𝑈) under budget constraint (𝑟 is given
disposable income) and choose the mix of final energy (𝐹𝐸 - the energy bundle),
further split in fossil fuels (𝐹𝐹𝐸), energy carriers (𝐸𝐶) and clean energy forms (𝐶𝐹𝐸),
and non-energy inputs (𝑁𝐸). Consumers perceive emission costs depending on a
shadow carbon price (𝑐𝑝), termed as carbon value, which is the dual variable of an
emission cap (4), but they do not actually incur carbon payments (this corresponds to
the concept of carbon value, as opposed to carbon price). Energy carrier producers
(problem (2)) minimize production costs (𝐶) to meet given demand (𝐸𝐶) increased by
distribution losses (𝑙𝑜𝑠 denotes the rate of losses). They mix fossil fuels (𝐹𝐸𝐶) and
clean energy forms (𝐶𝐸𝐶) through a production function. They price energy carriers
(3) so as to recover fixed and variable costs also depending on market competition
regime; tariff setting is denoted by function ℋ, depending on production costs (𝐶) and
volume of demand (𝐸𝐶); this function is a complex financial sub-model in PRIMES.
Tariffs may (optionally) include passing through of carbon costs to consumer prices,
depending on carbon price (𝑐𝑝) which is the dual variable of (4), if producers are
assumed to incur carbon payments (e.g. ETS). Total emissions, depend on unit
emissions (𝑒𝐹𝐹𝐸 , 𝑒𝐹𝐸𝐶 ) of fossil fuel consumption and have to be lower than a given

Page 11
 PRIMES 2018
cap (𝑐𝑎𝑝). Policy may impose a system-wide clean energy obligation (e.g. RES
obligation) expressed by (5) as share of gross final energy consumption (𝑟𝑒𝑠 is the
target and 𝑟𝑣 is the shadow price of the constraint, called RES value). Constraint (6)
introduces an energy saving (or efficiency) obligation, restricting final energy
consumption by a given upper bound (𝑠𝑎𝑣); shadow price to this constraint is 𝑒𝑣
called efficiency value.
In the formulation, ℵ is the utility function, ℱ is a production function. Their structure
define the substitution possibilities between fossil fuels, energy carriers and clean
energy forms at the consumers’ level, with ℊ being a production function mixing fossil
fuels and clean energy forms to produce energy carriers. ℘𝐹𝐹𝐸 and ℘𝐹𝐸𝐶 denote the
cost-quantity curves for fossil fuels addressed to consumers and energy carrier
producers, respectively. ℘𝐶𝐹𝐸 and ℘𝐶𝐸𝐶 are the cost-quantity curves of clean energy
forms used at consumer and producer levels respectively. All these functions are in
PRIMES complex sub-models and not analytical functions; similarly the pricing/tariff
equation is a complex sub-model. The cost-quantity curves (representing cost-supply
locus of a resource) apply in all demand and supply models to represent non-linear
resource constraints and price-responsiveness in relation to potentials. The concept
of resource cost curves apply on all possible potentials including energy efficiency,
RES, technology progress, storage, fuel supply, etc. The formulation below shows the
complex sub-models as simple functions for illustration purposes.
Max 𝑈 = ℵ[ℱ(𝐹𝐹𝐸, 𝐸𝐶, 𝐶𝐹𝐸), 𝑁𝐸]
𝐹𝐹𝐸,𝐸𝐶,𝐶𝐹𝐸,𝑁𝐸

Structural form of (1)

℘𝐹𝐹𝐸 (𝐹𝐹𝐸) ∙ 𝐹𝐹𝐸 + 𝑝𝐸𝐶 ∙ 𝐸𝐶 + ℘𝐶𝐹𝐸 (𝐶𝐹𝐸) ∙ 𝐶𝐹𝐸 + 𝑝𝑁𝐸 ∙ 𝑁𝐸 ≤ 𝑟
PRIMES model [
formulation
Min 𝐶 = ℘𝐹𝐸𝐶 (𝐹𝐸𝐶) ∙ 𝐹𝐸𝐶 + ℘𝐶𝐸𝐶 (𝐶𝐸𝐶) ∙ 𝐶𝐸𝐶 + 𝑐𝑝 ∙ 𝑒𝐹𝐸𝐶 ∙ 𝐹𝐸𝐶
𝐹𝐸𝐶,𝐶𝐸𝐶

(2) 𝐸𝐶
ℊ(𝐹𝐸𝐶, 𝐶𝐸𝐶) ≥
1 − 𝑙𝑜𝑠
[

(3) 𝑝𝐸𝐶 = ℋ(𝐶, 𝐸𝐶)

(4) 𝑒𝐹𝐹𝐸 ∙ 𝐹𝐹𝐸 + 𝑒𝐹𝐸𝐶 ∙ 𝐹𝐸𝐶 ≤ 𝑐𝑎𝑝 ⊥ 𝑐𝑝

𝐸𝐶
(5) 𝐶𝐹𝐸 + 𝐶𝐸𝐶 ≥ 𝑟𝑒𝑠 ∙ (𝐹𝐹𝐸 + + 𝐶𝐹𝐸) ⊥ 𝑟𝑣
(1 − 𝑙𝑜𝑠)

(6) ℱ(𝐹𝐹𝐸, 𝐸𝐶, 𝐶𝐹𝐸) ≤ 𝑠𝑎𝑣 ⊥ 𝑒𝑣

We firstly transform the optimisation problems (1) and (2) into equivalent mixed-
complementarity problems, to solve the EPEC problem. We take the derivatives of
Lagrange functions assuming that both demanders and suppliers see shadow prices
associated to system-wide constraints and that demanders are price takers. The
transformed problem is as follows (𝜆𝑢 is the marginal utility of income and 𝜆𝑐 is the
marginal cost of energy carrier production):

Page 12
 PRIMES 2018
𝜕℘𝐹𝐹𝐸
Mixed (6) 𝜆𝑢 ∙ ( ∙ 𝐹𝐹𝐸 + ℘𝐹𝐹𝐸 ) ⊥ 𝐹𝐹𝐸 ≥ 0
𝜕𝐹𝐹𝐸
complementarity 𝜕ℱ 𝜕ℵ 𝜕ℱ
+𝑐𝑝 ∙ 𝑒𝐹𝐹𝐸 + 𝑟𝑣 ∙ 𝑟𝑒𝑠 + 𝑒𝑣 ∙ + 𝑐𝑝 ∙ 𝑒𝐹𝐹𝐸 ≥ ∙
form of PRIMES 𝜕𝐹𝐹𝐸 𝜕ℱ 𝜕𝐹𝐹𝐸
model formulation 𝜕ℋ 𝑟𝑒𝑠 𝜕ℱ 𝜕ℵ 𝜕ℱ
(7) 𝜆𝑢 ∙ + 𝑟𝑣 ∙ + 𝑒𝑣 ∙ ≥ ∙ ⊥ 𝐸𝐶 ≥ 0
𝜕𝐸𝐶 1 − 𝑙𝑜𝑠 𝜕𝐹𝐹𝐸 𝜕ℱ 𝜕𝐸𝐶

𝜕℘𝐶𝐹𝐸 𝜕ℵ 𝜕ℱ
(8) 𝜆𝑢 ∙ ( ∙ 𝐶𝐹𝐸 + ℘𝐶𝐹𝐸 ) + 𝑟𝑣 ∙ (𝑟𝑒𝑠 − 1) + 𝑒𝑣 ∙ 𝑠𝑎𝑣 ≥ ∙ ⊥ 𝐶𝐹𝐸 ≥ 0
𝜕𝐶𝐹𝐸 𝜕ℱ 𝜕𝐶𝐹𝐸

𝜕ℵ
(9) 𝜆𝑢 ∙ 𝑝𝑁𝐸 ≥ ⊥ 𝑁𝐸 ≥ 0
𝜕𝑁𝐸

(10) 𝑟 ≥ ℘𝐹𝐹𝐸 (𝐹𝐹𝐸) ∙ 𝐹𝐹𝐸 + 𝑝𝐸𝐶 ∙ 𝐸𝐶 + ℘𝐶𝐹𝐸 (𝐶𝐹𝐸) ∙ 𝐶𝐹𝐸 + 𝑝𝑁𝐸 ∙ 𝑁𝐸 ⊥ 𝜆𝑢 ≥ 0

𝜕℘𝐹𝐸𝐶 𝜕ℊ
(11) ∙ 𝐹𝐸𝐶 + ℘𝐹𝐸𝐶 + 𝑐𝑝 ∙ 𝑒𝐹𝐸𝐶 ≥ 𝜆𝑐 ∙ ⊥ 𝐹𝐸𝐶 ≥ 0
𝜕𝐹𝐸𝐶 𝜕𝐹𝐸𝐶

𝜕℘𝐶𝐸𝐶 𝜕ℊ
(12) ∙ 𝐶𝐸𝐶 + ℘𝐶𝐸𝐶 − 𝑟𝑣 ≥ 𝜆𝑐 ∙ ⊥ 𝐶𝐸𝐶 ≥ 0
𝜕𝐶𝐸𝐶 𝜕𝐶𝐸𝐶

(13) ℊ(𝐹𝐸𝐶, 𝐶𝐸𝐶) ≥ 𝐸𝐶/(1 − 𝑙𝑜𝑠) ⊥ 𝜆𝑐 ≥ 0

(14) 𝑝𝐸𝐶 = ℋ(𝐶, 𝐸𝐶) ⊥ 𝑝𝐸𝐶 𝑓𝑟𝑒𝑒

(15) 𝑐𝑎𝑝 ≥ 𝑒𝐹𝐹𝐸 ∙ 𝐹𝐹𝐸 + 𝑒𝐹𝐸𝐶 ∙ 𝐹𝐸𝐶 ⊥ 𝑐𝑝 ≥ 0

(16) 𝐶𝐹𝐸 + 𝐶𝐸𝐶 ≥ 𝑟𝑒𝑠 ∙ (𝐹𝐹𝐸 + 𝐸𝐶/(1 − 𝑙𝑜𝑠) + 𝐶𝐹𝐸) ⊥ 𝑟𝑣 ≥ 0

(17) 𝑠𝑎𝑣 ≥ ℱ(𝐹𝐹𝐸, 𝐸𝐶, 𝐶𝐹𝐸) ⊥ 𝑒𝑣 ≥ 0

The system of complementarity conditions (6) to (17) is representative of the entire

PRIMES model.

D.2. Using PRIMES to meet policy targets: illustration

Assuming usual convexity conditions for problems (1) and (2), the consumers exhaust
disposable income and producers exactly meet demand. Thus, conditions (10) and
(13) lead to equality and the associated multipliers are strictly positive. As the
system-wide targets (15 to 17) entail increasing costs for being met, maximising
welfare suggests that they are exactly met in equilibrium when the bounds 𝑐𝑎𝑝, 𝑟𝑒𝑠
and 𝑠𝑎𝑣 are sufficiently stringent. It is possible that some or even none of the target
constraints bind at equilibrium, in which case the associated shadow prices are zero.
As all cost-supply functions (℘) are monotonically increasing, and so consumers and
producers use all kinds of inputs at equilibrium. In such case, all conditions (6) to (14)
are equalities in the optimal solution and the associated unknown variables are
strictly positive. The optimum use of utility enabling inputs (see conditions 6 to 9) is
determined at a level where marginal utility is equal to marginal costs including
marginal impacts on system-wide targets. Similarly, inputs to energy carrier

Page 13
 PRIMES 2018
production are determined at a level where marginal productivity equals marginal
costs including impacts on system-wide targets. Thus, meeting the system-wide
targets implies shifting away from inputs implying largest marginal deviation from
targets.
When policies set stringent targets, the demand for fossil fuels decreases but also
fossil fuel prices tend to decrease (due to the increasing fuel cost-supply curve). At the
same time unit costs of clean resources tend to increase, since their cost-supply curve
indicates that their use approaches maximum potential and therefore increased
marginal costs occur. The gradients of the cost-supply (or cost-quantity) influence
consumers and producers in their optimising behaviour.
Electricity prices are set through (14) at a level sufficient to recover all costs.
According to (10) the consumers do not pay directly for carbon emissions (unless
carbon pricing is implemented through cap and trade or via a tax), but they do take
into account shadow carbon prices in the choice of input mix charges on fossil fuels,
through (8), to determine their energy mix. They indirectly incur additional costs and
the purchasing power of income decreases; hence utility level decreases, when the
emission constraint (15) is binding. To compensate for this utility loss, additional
income would be necessitated, which correspond to valuation of disutility costs.
Similarly, consumers and producers are incited to meet the renewables and/or the
efficiency obligation as the renewables and efficiency values influence their
optimising behaviour in (6), (7), (8) for consumers and in (12) for producers. When
constraints (16) and (17) are binding, the renewable and efficiency values are non-
zero (positive) and the input mix is influenced both for consumers and producers, and
so indirectly costs increase.
When the policy constraints entail lower marginal costs in production of energy
carriers than in final consumption, then the consumers will tend to use more energy
carriers to the detriment of fossil fuels. This holds true if the change in energy carrier
price (𝑝𝐸𝐶 ) driven by carbon price is lower than the increase of marginal cost of clean
energy forms (℘𝐶𝐹𝐸 ) used directly by final consumers. An example is growing
electrification of demand.
Energy efficiency improvement is reflected also through substitution between the
energy bundle and the non-energy input to utility; example are more efficient use of
materials, change in habits, use of materials and equipment to increase efficiency of
building structures and factories and more efficiency in mobility. If substitution to
non-energy is less costly than substitution within the energy bundle at the final
energy demand level, then energy savings dominate and so the decarbonisation
possibilities in energy carrier production are of less importance. Conversely, if
substitution within the energy bundle is flexible enough and if emission reduction in
energy carrier production is flexible, then the energy carrier gets a higher share in
final energy demand and helps achieving lower emissions. Such a case occurs mostly
when time lags are sufficient to allow for renewing the capital stock in energy carrier
production; in the short term the impossibility to renew the capital stock imply low
adaptation flexibility in energy carrier production. The absence of flexibility in the
substitution between energy and non-energy at final demand level may lead to very
high compliance costs, as employing only substitutions within the final energy bundle
and within energy carrier production may imply high use of clean resources entailing
high nonlinear costs.

Page 14
 PRIMES 2018

Page 15
 PRIMES 2018

Page 16
 PRIMES 2018

E. Policy focus
PRIMES includes a rich representation of policy instruments and measures.
Modelling of Based on long experience with using PRIMES in major policy analysis and
policy measures: impact assessment studies of the European Commission, national governments
and industrial institutions, detailed mechanisms have been built in the model
 Wide coverage to represent a large variety of policy measures and regulations. Scenario
of policies construction assumptions about the inclusion of policies can be made in close
collaboration with the authority getting the modelling service because the
 Targets
modelling detail is high allowing for mirroring policies close to reality.
 Policies inducing The policy instruments classified in groups are as follows:
effects on costs
and prices Targets: they can be directly included in the model at various level, by sector,
by country, and EU-wide; they may concern emissions, renewables, energy
 Market-focusing efficiency, security of supply, fossil fuel independence, and others.
instruments Performance against targets derives from projection data. The PRIMES
reporting facility includes calculation of indicators according to regulations
 Regulations (e.g. RES shares).

 Standards Price or cost driving policies:

 Infrastructure  Taxation is exogenous and follows the level of detail of regulations, being
specific for fuels, sectors and countries. The data draw from the EU taxation
 Complex settings directives. Additional information determine values for subsidies and other
enabling higher forms of state supports.
effectiveness of  Cap and trade mechanisms and tradable certificate systems, including
structural Emission Trading Scheme, green and white certificates; the model
changes represents a variety of regimes and regulations, including grandfathering
and auctioning with different regulations by sector, and can handle floor
 Policy measures and cap prices as well as various assumptions about allowances and their
or targets can be composition. Trade of certificates or allowances can be handled over the
specified by EU or by country (or other grouping of countries) and also over time
sector, by including consideration of influence of foresight and risk-related
country or EU- behaviours
wide  Feed-in tariffs and other renewable support schemes: treated in great
detail in PRIMES including historical data and projection of consequences
over time; inclusion of possible budget constraints and modelling of
individual project developments on RES based on project-based financing
depending on support schemes totally or partially and the eventual
involvement of the RES project in the market.

Page 17
 PRIMES 2018
 Institutional mechanisms and regulations that may induce lower interest rates
Combination of
and lower perception of risks by individual investors; largely applied for
policies: modelling energy efficiency policies and other policies addressed to numerous
individuals.
 Market-oriented
 Contract for differences and purchasing agreements backed by the state aiming at
instruments can securing return on investment
be combined  Regulations and policies that address market failures and/or enable tapping on
with bottom-up positive externalities (e.g. technology progress) which induce reduction of cost
command and elements (technology costs) and improve perception by consumers leading to
control lower subjective cost components.
measures
Regulations on standards and command-and-control measures: they are explicit in the
model and depending on specification they are showing to eliminate certain
 Cap and Trade
technologies or options in the menu in technology choices in various sectors modelled
certificate
systems are  Eco-design standards in detail
modelled in  Best Available Technology regulations
detail and  Emission standards or efficiency standards on vehicles and other transport means
various regimes  Large combustion plant directives
can be mirrored  Emission performance standards
 Energy performance standards
 Policies having  Reliability and reserve standards (power and gas sectors)
indirect effects  Policies regarding permitting power plant technologies at national level, for
example regarding nuclear, CCS etc., including constraints applicable to new site
on risk and cost
development or expansion in existing sites. Also, policies regarding possibility of
perception of extension of lifetime of power plants (e.g.. nuclear) and retrofitting (e.g. to comply
actors, or with emission regulation)
inducing higher
Infrastructure policies and development plans in various sectors can vary in scenario
technology
assumptions and influence possibilities of technology deployment and system costs.
progress and
Coverage for infrastructure:
uptake
 Power interconnectors among countries, including expansion to remote areas for
 Focus on RES development purposes, and different options about management and
infrastructure allocation of capacities
development  Power grids and smart systems within countries, which are not spatially
represented but only through reduced-form cost-possibility curves in which
and its
parameters mirror development plans with influences on future technology
influences on
development (for RES, highly distributed generation, metering, demand response,
technology etc.)
possibilities and  Gas transport, LNG, storage and liquefaction infrastructure
on structural  Refuelling and recharging infrastructure in all transport modes
changes  CO2 transport and storage infrastructure

Page 18
 PRIMES 2018
Complex design of  Transport infrastructure parameters influence mobility and modal shifts but
policy scenarios: modelling does not include spatial information (limited to urban, semi-urban and
inter-urban)
 Explicit focus on  Hydrogen transport and distribution infrastructure (reduced form spatial
technologies, modelling)
systems and  Heat-steam district heating infrastructure (no spatial modelling)
infrastructure Enabling settings: direct policies as mentioned above or other policies (e.g. R&D)
combined with combined together may induce effects on technology costs or on perceived costs and
market risk factors or on actors’ behaviours thus enabling faster uptake of advanced or
functioning cleaner technologies thus making possible structural changes to happen in various
influenced by sectors. Examples are ambitious renovations of buildings and houses, electrification in
measures transport, development of alternative fuels, supply of new generation bio-energy
commodities, etc. The assumptions about enabling settings mainly influence
 Assumption of perception of costs, technology uptake and technology progress.
background ETS market simulation is explicit in PRIMES. However, the projections based on
policies enabling PRIMES are compatible with the 5-year time resolution of the model and the model
technology, algorithm only approximates the arbitration of allowances holders over time.
options and Nonetheless, PRIMES can handle multi-target analysis, for example, simultaneously for
ETS, non-ETS, RES and energy efficiency, where the aim is to determine optimal
resource
distribution of achievements (targets) by sector and by country. PRIMES has
availability for
successfully provides results for that purpose in the preparation of the 2020 Energy
energy purposes and Climate Policy Package (2007-2008) and recently for the 2030 Policy Analysis
(2013).
 PRIMES
reporting and Detailed reporting and ex-post calculations: to support impact assessment studies
ex-post PRIMES provides detailed reports of scenario projections. The reports calculate cost
indicators (with various levels of detail distinguishing between cost components and
calculations, as
sectors), as well as for numerous other policy-relevant indicators. Topics covered
well as through include environment, security of supply and externalities (e.g. noise and accidents in
linkages with transport). Thus, the model provide elements and projections to support cost-benefit
other models, analysis studies, which are the essential components of impact assessments. When
provides rich PRIMES links with the macroeconomic model GEM-E3, the coverage of projection data
numerical for the purposes of cost-benefit evaluations is more complete and comprehensive.
projection data Similarly, linkages with GAINS (from IIASA) provide wider coverage of cost-benefit
projections regarding atmospheric pollution, health effects, etc.
for feeding cost-
benefit and
impact
assessment
studies

Page 19
 PRIMES 2018

F. Typical Inputs and Outputs of PRIMES

Inputs  GDP and economic growth per sector (many sectors)
 World energy supply outlook – world prices of fossil fuels
 Taxes and subsidies
 Interest rates, risk premiums, etc.
 Environmental policies and constraints
 Technical and economic characteristics of future energy technologies
 Energy consumption habits, parameters about comfort, rational use of
energy and savings, energy efficiency potential
 Parameters of supply curves for primary energy, potential of sites for
new plants especially regarding power generation sites, renewables
potential per source type, etc.

Outputs  Detailed energy balances (EUROSTAT format)

 Detailed demand projections by sector including end-use services,
equipment and energy savings
 Detailed balance for electricity and steam/heat, including generation by
power plants, storage and system operation
 Production of fuels (conventional and new, including biomass
feedstock)
 Investment in all sectors, demand and supply, technology
developments, vintages
 Transport activity, modes/means and vehicles
 Association of energy use and activities
 Energy costs, prices and investment expenses per sector and overall
 CO2 Emissions from energy combustion and industrial processes
 Emissions of atmospheric pollutants
 Policy Assessment Indicators (e.g. import dependence ratio, RES ratios,
CHP ratios, efficiency indices, etc.)

Coverage of  38 European countries (individual projections)

PRIMES inputs  2010 – 2070 by 5-years steps
and outputs  Trade of electricity, gas and other fuels between the European countries
and with the rest of the World

Page 20
 PRIMES 2018

G.Comparison to other models

What PRIMES can do The distinctive feature of PRIMES is the combination of micro-economic
foundations with engineering at a high level of detail, compatible with a long-
term time scale and sectoral detail of available statistics for Europe

Designed to provide long-term energy system projections and system

restructuring up to 2050, in both the demand and the supply sides. Projections
include detailed energy balances, structure of demand by sector, structure of
power system and other fuel supplies, investment and technology uptake,
costs per sector, overall costs, consumer prices and certificate prices (incl.
ETS) if applicable, emissions, overall system costs and investment.

Impact assessment of specific energy and environment policies, applied at

Member State or EU level, including price signals, such as taxation, subsidies,
ETS, technology promoting policies, RES supporting policies, efficiency
promoting policies, environmental policies

The linked model system PRIMES, GEM-E3 and IIASA’s GAINS (for non-CO2
gases and air quality) perform energy-economy-environment policy analysis in
a closed-loop

What PRIMES No forecasting but scenario projections. PRIMES is not an econometric model.
cannot do Cannot perform closed-loop energy-economy equilibrium analysis, unless
linked with a macroeconomic model such as GEM-E3.

PRIMES has more limited resolution than engineering electricity, refinery and
gas models dedicated to simulating system operation in detail. Although rich in
sectoral disaggregation, PRIMES has limitations due to the concept of
representative consumer per sector, as it does not fully capture the
heterogeneity of consumer types and sizes.

PRIMES lacks spatial information and representation (at a level below that of
countries) and so it does not fully capture issues about retail infrastructure for
fuels and electricity distribution, except for electricity and gas flows over a
country-to-country based grid infrastructure, which is well represented in the
model

PRIMES is an empirical numerical model with emphasis on sectoral and

country specific detail; it has a very large size and so some compromises were
necessary to limit computer time at reasonable levels.

PRIMES differ from overall optimization energy models, qualified by some as

bottom-up approaches, as for example MARKAL, TIMES, EFOM. Such models

Page 21
 PRIMES 2018
formulate a single, overall mathematical programming problem, do not include
explicit energy price formation and have no or simple aggregate
representation of energy demand. PRIMES formulates separate objective
functions per energy agent, simulates in detail the formation of energy prices
and represents in detail energy demand, as well as energy supply.

PRIMES differ from econometric-type energy models, such as POLES, MIDAS

and the IEA’s World Energy Model. These models use reduced-form equations
that relate in a direct way explanatory variables (such as prices, GDP etc.) on
energy demand and supply. These models have weak representations of useful
energy demand formation. They are usually poor in representing in detail
capital vintages and technology deployment in energy supply sectors and lack
engineering evidence, as for example the operation of interconnected grids and
detailed dispatching.

PRIMES is a partial equilibrium model as opposed to general equilibrium

models, such as GEM-E3.

Obviously, PRIMES differs substantially from accounting-type models, which

usually focus on specific sectors, such as MEDEE, MAEDS (on energy demand),
GREEN-X (renewables), BIOTRANS (biofuels), etc.

The distinguishing feature of PRIMES is the representation of each sector

separately by following microeconomic foundations of energy demand or
supply behaviour and the representation of market clearing through energy
prices. Similar models developed in the USA, include PIES, IFFS and mainly the
NEMS model, which is currently the main model of USA DOE/EIA.

These models are qualified as generalized equilibrium models because they

formulate the behavioural conditions for economic agents and combine a
variety of mathematical formulations in the sub-models, represent different
market clearing regimes. These models are also qualified as hybrid models
because they combine engineering-orientation with economic market-driven
representations.

Page 22
 PRIMES 2018

H. Overview of PRIMES model resolution

REGIONS: The PRIMES model is operational for all EU28 individual member-states and for the Western
Balkans countries (Albania, Bosnia-Herzegovina, FYR of Macedonia, Serbia, Kosovo and
Montenegro), the EFTA countries (Switzerland, Norway, Iceland) and Turkey. It
projects also the flows of electricity and gas among all countries. A simple version of
the model runs on data for 11 North African and Middle East countries.
FUEL TYPES: PRIMES projects energy demand and supply balances distinctly for 45 energy
commodities and forms. The list is:
 coal, lignite, coke, peat and other solid fuels;
 crude-oil, feedstock oil, residual fuel oil, diesel oil, liquefied petroleum gas,
kerosene, gasoline, naphtha, other oil products; bio-fuels (several types);
 natural and derived gasses (blast furnace, coke oven and gas works, as week as oil
and solids gasification outputs);
 thermal solar (active, high enthalpy and low enthalpy), geothermal low and high
enthalpy;
 steam/heat (industrial and distributed heat);
 electricity, nuclear energy;
 biomass and waste (5 bio-energy types and several feedstock types);
 solar PV electricity, solar thermal electricity, wind onshore, wind offshore, hydro
lakes, hydro run of river, tidal and wave energy
 Hydrogen, e-fuels (synthetic gas and liquids).
The model projects volumes and prices by fuel type and by sector.
RESIDENTIAL: The residential sector includes 54 building types by age, location, and building type. The
model includes 29 space heating and cooling equipment types, water heating and
cooking. The electric appliances (several categories) for non-heating purposes reflect
technology vintage dynamics, eco-design regulations and follow stock-flow relations.
There is no distinction between rented and owned dwellings.
SERVICES & AGRICULTURE: The model distinguishes between two commercial sectors and one public
sector, further split into 8 subsectors. At the level of each sub-sector, the model
calculates energy services (useful energy), which are further subdivided in energy uses
(several types) defined according to the pattern of technology. Service buildings are
also categorised by age. The model includes in total more than 30 end-use technology
types.
INDUSTRY: The industrial model formulates 10 industrial sectors separately and 31 subsectors, namely
iron and steel, nonferrous (several sectors), chemicals subdivided in basic chemicals,
petrochemicals, fertilizers, cosmetics/pharmaceuticals, non-metallic minerals
subdivided in cement, ceramics, glass and other building materials, paper and pulp
subdivided in pulp, paper and printing, food drink tobacco, engineering, textiles, other
industries and non-energy uses of energy products. For each sector different sub-
processes are defined (in total about 30 sub-sectors, including focus on materials and
on recycling; sectors are subdivided in sub-sectors based on whether processing is
based on primary or scrap feedstock). At the level of each sub-sector a number of
different energy uses are represented (the model includes in total about 235 types of
energy process technologies).
TRANSPORT: The transport sector distinguishes passenger transport and goods transport as separate
sectors. They are further subdivided in sub-sectors according to the transport mode

Page 23
 PRIMES 2018
and mean (public, private, road, rail, maritime, air, etc.). At the level of the sub-sectors,
the model structure defines several vehicle types and categories including distinction
by size by purpose or trip type and by technology type. Within modes like road
transport there is a further subdivision, e.g. for road passenger transportation the
model distinguishes between public road transport, metro, other rail, fast trains,
motorcycles and many types of private cars. The model considers several alternative
technologies and fuels for each transport mean. The model also projects activity by
typical area (urban, semi-urban and inter-urban) and by trip type. In total, the model
includes 15 transport modes, 103 vehicle types for road and non-road transport, 4
stylized geographic areas, distinction between peak and off-peak and 3 freight
categories.
ELECTRICITY AND STEAM PRODUCTION: Very detailed model including 72 different plant types per
country for the existing thermal plant types; 150 different plant types per country for
the new thermal plants; 3 different plant types per country for the existing reservoir
plants; 30 different plant types per country for the intermittent plants. In total the
database includes approx. 13000 power plants. Chronological load curves for
electricity and steam/heat distributed, 3 voltage types for the grid, interconnecting
European system in detail (individually for all interconnectors, present and future,
including ENTSOe development plans), network capacity and electric characteristics of
interconnectors. The power/steam model represents three stylised activities with
distinction between utilities, industrial production and highly distributed scale as well
as for self-power generation. Cogeneration of power and steam (12 generic
technologies), district heating, industrial boilers by sector, and distinction between
plants in industrial sites and merchant CHP.
NATURAL GAS: Very detailed sub-model covering regional supply detail (Europe, Russia, CIS countries
Middle Africa, North Sea, China, India for pipeline gas and global market for LNG).
Detailed representation of gas infrastructure (field production facilities, pipelines, LNG
Terminals, Gas Storage, Liquefaction Plants).
BIOMASS SUPPLY: Very detailed sub-model covering supply of biomass and waste energies including a
wide variety of feedstock types and transformation processes into bio-energy
commodities including bio-refineries. The model covers several land categories,
resources (crops, forestry, aquatic biomass and wastes) and of more than 35
transformation processes. Covers life-cycle calculations.
REFINERIES: Simple oil refinery type with typical refinery structure defined at the level of each
country; 5 typical refining units (cracking, reforming etc.)
HYDROGEN: Detailed hydrogen production and transportation sub-model with 18 H2 production
technologies, 8 H2 transport/distribution means and several types of H2 using
equipment.
PRIMARY FOSSIL FUEL PRODUCTION: Simple Cost – Supply curves limited by available resources
covering all primary energy extraction activities including conversions to briquetting,
liquefaction and gasification.
EMISSIONS: CO2 emissions from energy combustion, process-related in industry, Atmospheric
Pollutants (SO2, NOx, PM, VOC), ETS and non-ETS split, and non CO2 GHG abatement
cost curves provided by GAINS (IIASA).

Page 24
 PRIMES 2018

I. Data Sources and Model Linkages

I.1. Data sources
EUROSTAT: Energy Balance sheets, Energy prices (complemented by other sources, such IEA),
Macroeconomic and sectoral activity data (PRIMES sectors correspond to NACE 3-digit
classification), Population data and projection, Physical activity data (complemented by other
sources), CHP surveys, CO2 emission factors (sectoral and reference approaches) and EU ETS
registry for allocating emissions between ETS and non ETS

TECHNOLOGY DATABASES: MURE, ICARUS, ODYSEE – demand sectors, VGB (power technology costs),
TECHPOL – supply sector technologies, NEMS model database, IPPC BAT Technologies IPTS

OTHER DATABASES: District heating surveys, buildings and houses statistics and surveys (various sources),
IDEES, BSO, BPIE,

POWER PLANT INVENTORY: ESAP SA and PLATTS

RES POTENTIAL: ECN, DLR and EURObserver

NETWORK INFRASTRUCTURE: ENTSOE, ENTSOG, GIE, TEN-T (transport infrastructure)

I.2. Model Linkages

GEM-E3: Linkage to GEM-E3 to take projections of activity by sector/country and GDP and to send energy
projections to GEM-E3 in order to carry out closed-loop macroeconomic impact assessment
studies

PROMETHEUS OR POLES: Linkage to these global energy models to take projections of world fossil fuel
prices

GAINS: Linkage to GAINS to take marginal abatement cost curves for non-CO2 greenhouse gases and to
convey energy projections to GAINS in order to evaluate impacts on atmospheric pollution

CAPRI, GLOBIOM: Linkage to send to these models detailed biomass supply projections in order to evaluate
land use and LULUCF impacts

TRIMODE: Linkage to a spatial transport flow model to take activity projections for mobility in order to
calibrate a reference projection (PRIMES provides its own activity projection in scenarios)

MODELS CALCULATING POTENTIALS: PRIMES uses detailed bottom-up information on energy efficiency
and renewable potential (databases and models including DLR, GREN-X and several others)

Page 25
 PRIMES 2018

Macroeconomic/sectoral activity
GEM-E3 model
Transport activity
Flows
SCENES or
TRANSTOOLS

Energy World energy

Oil, gas, coal prices
demand-supply-prices,
emissions and POLES or Prometheus model
investment
PRIMES
model
EU power plants – Platts
Technologies (TechPol,VGB)

EU refineries - IFP

Renewables potential DLR, ECN,

 Air Quality and non CO2 GHG Observer
emissions – IIASA - GAINS
model
 Land and LULUCF impacts Energy efficiency Fraunhofer, Wuppertal,
(CAPRI, GLOBIOM) ODYSEE, MURE databases

Page 26
 PRIMES 2018

J. Stationary Energy Demand Sub-models

J.1. General Methodology
For each energy demand sector a representative decision-making agent
Main Features
operates, who optimizes an economic objective function. For households and
passenger transport, the model formulates a utility maximisation problem. The
 Energy demand
model uses a profit maximisation (or cost minimisation) function for
based on individual
industrial, tertiary and freight transport sectors. The decision on fuel and
optimizing technology mix follows a nested budget allocation problem.
behaviour of
representative Firstly useful energy demand (services from energy such as temperature in a
agents by sector house, lighting, industrial production, etc.) is determined at a level of a sector.
At the upper level of the nesting, energy is a production factor or a utility
 Choices follow a tree providing factor and competes with non-energy inputs. At this level a
with useful energy on macroeconomic econometrically estimated function is used which combines
top and choice of energy and non-energy inputs and considers saturation dynamics. Saturation
fuels at the bottom depends on income for households and the saturation factor exhibits a sigmoid
curve which indicates income elasticity of energy above one if useful energy at
 Detailed low levels (less developed countries) and elasticity values lower than one (and
decomposition in decreasing) when useful levels are high.
several processes and
Useful energy, as derived, decomposes into uses and processes (e.g. space
uses of energy
heating, water heating, motor drives, industrial processes, etc.). The separation
in uses and processes follows a tree structure which is formulated
 Focus on materials in
mathematically so as optionally to allow either for complementarity or
industry and explicit
substitutable relationships among uses/processes. For example to produce a
modelling of
certain product, the model activates a certain chain of process flows: in this
recycling
case, they are complementary with each other. However, it may be that the
 Explicit product can equally go through electro-processing or thermal processing in
which case the processes are substitutable to each other. For some sectors, the
representation of
model distinguishes between sub-sectors in order to get a more accurate
equipment using
representation of the stylised agent. For industrial sectors, the model puts
fuels and distinction
emphasis on materials and recycling and so it distinguishes between sub-
between purchasing
sectors, which involve basic processing (e.g. integrated steelwork, clinker in
of fuels and self-
cement, primary aluminium, etc.) and sub-sectors which use recycled and
production
scrap material. The possible substitutions between such sub-sectors is
endogenous, and depend on prices, policy measures, macroeconomic demand
factors and maximum potential of recycling possibilities, which are captured
through increasing cost-potential curves. The choice of elasticity values and
specific functional forms expresses the a priori considerations about

Page 27
 PRIMES 2018

Main Features complementarity and substitutability structure among sub-

sectors/uses/processes in each final energy demand sector.
 Equilibrium between
Useful energy requirements at the level of sub-sectors/uses/processes (e.g.
energy demand and
space heating, air conditioning, lighting, motive power, etc.) links to
(self) supply is consumption of final energy.
modelled in each
sector The representative agent in each sector or sub-sector makes choices among
fuels, technologies and energy savings to minimize costs in meeting the useful
 Useful energy energy requirements. The formulation includes the possibility of choice
demand depends on between purchasing ready-to-use fuels or energy carriers and self-producing
energy energy where this is possible. Examples are cogeneration versus district
saving/efficiency heating, etc.
investment and
The least cost choice is dynamic and involves endogenous choice of equipment
behaviour; tapping (vintages, technologies and learning), endogenous investment in energy
on saving potential efficiency (savings), endogenous purchase of associated energy carriers and
induces non-linearly fuels (demander is price taker). These are capital budgeting decisions which
increasing costs may involve trade-offs between upfront costs and variable-running costs.
Capital decisions use weighted average cost of capital (WACC) and subjective
 Self-supply derives
discount rates to annualise (levelized) costs to compare with variable-running
from dynamic least costs, which by definition are annual. The model for all demand sectors
cost optimization dynamically tracks capital accumulation with endogenous investment, tracking
which involves of vintage characteristics and endogenous premature scrapping.
technology choice
depending on The aim of the modelling is to mimic decisions by individuals as realistically as
perceived costs and possible. Subjective discount rates and business WACC include risk premium
on subjective factors, which reflect opportunity costs of drawing funds by the private sector.
They also reflect uncertainty, lack of information and probably limited access
discount rates
to capital markets. For this reason, the model relates the individual discount
 Policies are explicitly rates with a policy context, in order to mirror how certain policy instruments
represented and may reduce uncertainties or decrease financing costs in order to make
influence choices, economic decisions for technologies with high upfront costs. To mimic reality,
technology the model also includes several non-engineering cost facts which represent
technical uncertainty, risk of high costs of maintenance in case of not-yet
economics,
mature technologies, easiness of technology application, easiness to comply
perception of costs
with permits and regulations, etc. The terminology used is that the user sees
and discount rates
perceived cost values for technologies and solutions where some of the cost
components can reduce over time as technology becomes commercially
mature. This is one of the ways for representing endogenous learning-by-doing
mechanisms in the model. Thus, decisions at each nesting level use equivalent
perceived costs to reflect actual costs, utility (e.g. comfort), and uncertainty

Page 28
 PRIMES 2018

Main Features and risk premiums. The decisions depend on policy measures, such as: taxes
and subsidies, promotion of new technologies (reducing perceived costs), and
 Detailed promotion of energy efficiency, including standards -e.g. CO2 regulations for
representation of passenger car or regulations on minimum performance of lighting, policies
various energy that ease financing, etc.
efficiency promoting For industrial energy demand, PRIMES follows a formulation that allows for
policy instruments full integration with macroeconomic production functions. Sectorial value
added derived from GEM-E3 projections (general equilibrium macroeconomic
 Heat pumps and
model), link to PRIMES measurement of activity in physical units. Substitutions
directly used
between energy and non-energy (capital) inputs is handled at the upper level
renewables are
of PRIMES nesting and can coordinate with GEM-E3 projections. A large
represented number of industrial processes (e.g. different for scrap or recycling processes
and for basic processing) as well as a mix of technologies and fuels, covering
 Several dwelling
the use of self-produced by-products (e.g. black liquor, blast furnace gas)
types and several
provides higher resolution of industrial processing in PRIMES than in GEM-E3.
services sectors
Energy savings possibilities follow engineering representations, including the
 Dynamic modelling possibilities of shifting towards more efficient process technologies.
of technologies Substitutions are possible between processes, energy forms, technologies and
tracking vintages energy savings. The adoption of technologies depends on standards, emission
and including constraints, pollution permits and is dynamic keeping track of technology
endogenous learning vintages and stock-flow investment. The actual lifetime of existing equipment
is endogenous driven by relative costs.
 Investment, use of
existing stock and The industrial model considers explicit ways of producing steam, for example
using boilers or CHP. The model distinguishes between boiler steam, CHP
possible premature
steam from onsite plants and distributed CHP steam. Interaction with Power
scrapping are
and Steam sub-model for industrial CHP and boilers is an integral part of the
endogenous
model. The choices at industrial scale consider steam-driven CHP and CHP
decisions
driven by electricity-market. The model has a database on onsite CHP, which
 The demand sub- are cogeneration units with no access to steam distribution. The official
statistics do not include these onsite plants. A special routine in the PRIMES
models are solved as
database combines Eurostat statistics on energy balances and CHP surveys,
non-linear mixed
isolates in the data the on-site CHP, and reconstitute inputs and outputs for
complementarity
such installations.
problems
PRIMES represents possible substitutions and energy efficiency at various
levels in the residential and tertiary sectors and includes special routines for
the building stock and its renovation. The model tracks the dynamics of the
building stock, split by categories, and formulates demolishment decision,
construction of new buildings and renovation with distinction of various

Page 29
 PRIMES 2018
Main Features degrees of renovation for energy saving purposes. These decisions derive from
economics and are simultaneous with the nested decisions on useful energy
 Efficiency obligations demand, fuel mix choice, equipment choice, and energy efficiency investment.
and white Rebound effects stemming from cost savings due to energy efficiency are
certificates are present and derive simultaneously with the rest of decisions. For example,
included useful energy demand may increase because of high energy efficiency gains.
The decisions related to buildings also depend on behavioural characteristics
 Standards influence and are influenced by perceived costs, subjective discount rates and prices.
the menu of Policy measures and instruments, and standards such as the building codes
technology choice influence the decisions. The model includes heat pumps and direct use of
renewables (biomass, solar, geothermal, etc.). The related decisions are
 Engineering details simultaneous with the rest of decisions, including the dynamic track of
for industry capture technology vintages.
processing of
Surveys have shown that the substitution possibilities and the energy
materials and choice
efficiency investment depend on the main pattern of space heating method,
between processing
which is a goof dimension to classify the various behavioural types. For this
technologies
purpose, the model includes a distinction of five dwelling types according to
space heating pattern; one of the categories group partly heated houses.
 Cogeneration versus
PRIMES also distinguishes agriculture and services sectors which are broken
industrial boilers are
down by sub-sector (e.g. market services, non-market services, trade); electric
closely linked to
appliances and lighting are separately treated in all sectors.
electricity and heat
markets; on-site CHP The following diagram illustrates the tree decomposition of each energy
is explicitly demand sector in sub-sectors, further in processes and in energy uses. A
represented technology operates at the level of an energy use and utilizes purchased
energy forms (fuels and electricity) or self-produces energy. The calculation
starts from activity or income, then it computes useful energy and then by
using technology equipment it meets useful energy by converting purchased or
self-produced energy forms (final energy). The mathematical formulation of
the nested decisions solves as a whole, including the least-cost choice of
technologies and fuels and the dynamic investment process.

The demand models solve as mixed complementarity problems, which

concatenate the individual optimization problems written in the form of Kuhn-
Tucker conditions.

Page 30
 PRIMES 2018

Summary of Link to Macro-Economy

Activity and Income Variable
energy
demand Sector - 1
e.g. Iron and Steel
Sector - 2
e.g. Residential
Sector - 3
e.g. Passenger transports
modelling
methodology Sub-Sector - 1
e.g. Central Boiler Dwellings
Sub-Sector - 2
e.g. Electric Heating Dwellings

Energy Use -1 Energy Use - 2

e.g. Space Heating e.g. Water Heating

TECHNOLOGY
Ordinary

Fuels

Fuels

Future
Technologies

The model evaluates consistently the potential of new technologies, by considering simultaneously four types
of mechanisms: a) economic optimality, b) dynamics, i.e. constraints from existing capacity, c) gradual market
penetration depending on relative costs and risk perception, d) endogenous technology learning and
commercial maturity.
The non-linear optimization per agent (sector) performs dynamically in a time forward direction with
foresight limited to 10 years. In a given period a set of lagged values up-dated dynamically by the single-period
optimization results reflect adaptive expectation over 10-years. Choices are constrained dynamically by the
existing energy-use equipment stock, which may change through investment while existing equipment can be
retired based on retirement rates, or by premature replacement decisions. Technology (energy equipment that
converts purchased energy to useful energy) and energy savings equipment (e.g. insulation) is considered to
evolve over time, and is categorized in vintages (generations) presenting different cost and performance
features.
The upper level functions which project useful energy demand (services provided by using energy or by saving
energy) are of econometric nature and are based on complex functional forms relating demand with
macroeconomic drivers so as to capture possible saturations, rebound effects and comfort depending on
income growth. The useful energy demand functions are dynamic and depend on evolution of unit cost of
energy services, which aggregate costs of equipment for operation and investment in various energy uses and
for saving energy. Investment enabling energy efficiency progress at useful energy level concerns improvement
of thermal integrity of houses and buildings, horizontal energy management systems in industry or offices, etc.
Such investments are determined together with useful energy demand to fully capture rebound effects and
depend on investment costs, energy prices, carbon prices and policies supporting or facilitating such
investments. Stock turnover dynamics, including for renovation, are explicit in the model. Costs related to
energy saving potentials are non-linear assuming exhaustible potentials and cost gradients increasing with
volumes of energy savings due to upper level investments. Discrete choice theory formulations capture
heterogeneous situations regarding house/building types and conditions. Heterogeneity also justifies the non-
linear costs but are difficult to represent analytically due to lack of statistics. The non-linear cost-saving
possibility curves are estimated using micro and bottom-up sources based on surveys and available databases.

Page 31
 PRIMES 2018

J.2. Industrial energy demand sub-models

In PRIMES, industry consists of ten main sectors, which split in 31 different
sub-sectors. Each sub-sector includes a series of industrial processes and
energy uses totalling 234 uses; additionally, 22 different fuel types are
available for the industrial sectors. Technologies are explicit, firstly at the
process level where different process types are included and secondly at the
levels of energy uses where technologies use different types of fuels. The
model distinguishes between low enthalpy heat and steam. Heat and steam can
be either self-produced using boilers or CHP or purchased from the steam or
heat distribution markets, which depends on other industries’ CHP or boilers.

PRIMES handles in
great detail energy
intensive
processing in Figure 1: Overview of the sectors and subsectors included in PRIMES industry
industry and the
ways of reducing
energy demand
and shifting away The structure of processes and uses in the industrial sector can be seen in the
from GHG figures at the end of this section. The current model version splits alumina
emissions
production from primary aluminium production (previously grouped into
one), clinker from cement production (particularly important, as clinker

Page 32
 PRIMES 2018
imports tend to increase over time) and includes a large list of sector-specific
processes. In particular, it further includes significant details for pyro-
metallurgy, fire refining, and electro-refining options used for the production
of non-ferrous metals. The database of the techno-economic data for the split
of process and technologies has been updated –for the last Reference scenario
published 2016- based on extensive literature research (IEA-ETSAP, industrial
surveys, etc.). The energy saving possibilities from technologies included in the
eco-design directive (e.g. air compressors, etc.) were verified, emerging
technologies such as gas and liquid membrane technologies for separation in
the chemical industry are included and are taken into account in the industrial
module of PRIMES.

The scope of the industrial demand sub-model of PRIMES is to represent

simultaneously:

 the mix of different industrial processes (e.g. different energy intensity for
scrap or recycling processes and for basic processing);
 the mix of technologies and fuels, including the use of self-produced by-
products (fuels) and renewable energy forms;
 the links to self-supply of energy forms (e.g. cogeneration of electricity-
steam, steam by boilers, use of by-products (fuels), heat recovery);
 the explicit and engineering-oriented representation of energy saving
possibilities;
 the satisfaction of constraints through emission abatement, pollution
permits and/or energy savings, and
 the rigidities of system change evolution because of existing capacities or
dynamic technical progress
 Possible substitutions between processes, energy forms, technologies and
energy savings
 CO2 capture and process related emissions are included in the model
Energy efficiency improvement in industry is linked to technology choices at process
and energy use levels, and in addition derives from direct investment on energy
savings; all options are fully included in the modelling.

 Direct investments in energy savings are modelled as horizontal energy

management systems and as specific interventions mainly for heat recovery.
The saving possibilities follow cost-quantity curves, which have limited
potential and non-linear increasing costs.
 Choice of process technologies and energy use technologies (equipment) involve
a variety of possibilities, which differ in upfront costs and in variable costs
depending on energy performance. Processing and energy using equipment
stock turnover is dynamic, keeping track of technology vintages; both the choice
of technologies as well as the stock turnover can be influenced by policies which
are represented in the modelling.

Page 33
 PRIMES 2018
Similarly, to the energy efficiency the model includes a variety of options for the
mitigation of CO2 emissions, these are linked to the choice of technologies and fuels
at process and energy use level. Obviously the emission savings deriving from
reduced fuel consumption thanks to energy efficiency measures are automatically
taken into account in the model.

 Fuel switching: the model allows for fuel switching towards e.g. gas or electricity
(for the latter see next bullet), where the processes allow to move from more
carbon intensive fuels to less carbon intensive fuels, such as switching from coal
to gas. The model allows also for the use of hydrogen where this is technically
feasible (e.g. co-firing of hydrogen in high temperature furnaces), as well as the
use of waste, biomass and e-fuels (hydrogen or methane, as well as “synthetic”
liquid fuels).
 Electrification of processes/uses: electrification often leads to energy savings, as
well as CO2 mitigation. The PRIMES model includes a variety of options for the
electrification of processes, including processes available today such as electric
boilers, electric arc furnaces and microwave heating, as well as processes which
are expected to be commercially available in the future such as mechanical
drives to replace steam drives or high temperature heat pumps.

Figure 2: Horizon of commercial deployment for electrification options in industry

The PRIMES model has been recently updated to include several options to allow for
deep decarbonisation of the industrial sector including:

Page 34
 PRIMES 2018
 Use recycled or renewable carbon instead of fossils as feedstock to produce
complex molecules in chemicals (electro-chemical reduction technologies);
 Negative emissions in carbon feedstock to polymers;
 Integration of low carbon solutions for combustion with process emissions;
 Industrial symbiosis – exchange to recycle carbon, syngas and others;
 Circularity: increase recycling of materials and products;
 Use electricity in separation, heat uses and low enthalpy heat electrification
(UV, infrared, microwave, induction, etc.);
 Direct reduction of iron ore for steel;
 New cement chemistries;
 Capture of CO2 and syngas from steel and cement (and other processes) to
reuse for recycled carbon feedstock;
 Efficient separation of CO2 from flue gas and process flows;
 Electricity for vapour recompression (electrical steam);
 Electricity-driven separation;
 Medium to High temperature heat pumps;
 Heat recovery;
 Recovery of low concentration compounds.

The representation and reduction options for CO2 from process emissions have
recently been improved in the PRIMES model. They are computed through
relationships driven by the physical production of the relevant industrial commodities
(e.g. cement) and the process type used for the production (e.g. direct ore reduction
vs. blast furnace) and by the fuel use (hydrogen co-firing in furnaces). Remaining
emissions may be reduced through Carbon Capture techniques, which apply on the
processing of industrial commodities. The representation includes capital and
variable costs of CCS, as well as electricity consumption associated with capture,
which adds up to total demand for electricity. The model allows for both storage and
utilization of the captured carbon emissions, as well as storage in chemical materials
(plastics).

Industrial Boilers and cogeneration (CHP) are covered in a separate module of

PRIMES, where they are included by sector to allow for representing the specificities
of industrial sectors in an accurate manner. The steam required by different
industries differs in quality (temperature-pressure ratios), implying that the boilers
are “calibrated” to the specific industry which requires them, making boilers specific

Page 35
 PRIMES 2018
for certain industries and not easily substitutable e.g. with supply through steam
distribution networks. The inclusion of boilers and CHP by industrial sector allows
for the representation of the heterogeneity and inertia of behaviour, both captured
through discrete choice modelling techniques. The boilers and cogeneration units
each have different fuel options which can be influenced by prices or through policies,
thus allowing for emission reductions; the model also allows for switching between
boilers and CHP to enhance overall system efficiency.

Page 36
 PRIMES 2018

Page 37
 PRIMES 2018

Page 38
 PRIMES 2018

Page 39
 PRIMES 2018

Page 40
 PRIMES 2018

Page 41
 PRIMES 2018

Page 42
 PRIMES 2018

J.3. Steam and heat generation

The current version of PRIMES splits the modelling of steam and heat generation into
sub-models covering boilers and cogeneration for each industrial sector, district
heating including heat extraction from cogeneration and separately the rest of the
power market. The three main modelling blocks communicate with each other by
exchanging flows (power, heat and steam) and by exchanging money transfers based
on prices for the respective commodities. The split of the models is beneficial for
representing specificities of industrial sectors in a more accurate manner. It is also
beneficial for introducing heterogeneity and inertia of behaviours, both captured
through discrete choice modelling techniques. The enhanced model has better
stability than before in the simulation of cogeneration which is more dependent on
industrial sector circumstances than on conditions prevailing in the power market.

The split of the model, allows the introduction of higher resolution in terms of
technologies. For example, in district heating the model includes electric boiler, heat
pumps and in detail the various biomass and waste technologies and feedstock types,
for example with distinction between landfill, sewage, solid waste, wood and wood
waste, and biogas facilities.

Distributed steam is a small fraction of total steam consumption; however, the

statistics provide very poor steam generation information by sector. Even the
cogeneration surveys, carried out by Eurostat, discontinued after 2010 and anyway
they were not available by year in full detail. The PRIMES model has a method for this
disaggregation of steam heat production, which will be maintained until better data is
available. The enhanced method combines information from the CHP surveys, the
power plant inventory (as further processed to identify sector origins), engineering
information on stylised process flow by type of industry (30 industries in total) and
the power generation statistics which include the fuels corresponding to the
electricity generation from cogeneration, nonetheless without split between pure
electric and CHP generation. The data for the time series from 2000 until 2015 has
been constructed through algorithms for matching and consistency checking, which
generally follow cross-entropy methodology. The engineering templates of process
flows draw on a large set of reports from JRC (IPPC directive), from industrial
associations, specialised handbooks, IEA technology reports and the analytical
surveys by sector, which are available in the USA (EIA).

In industrial sectors, practice suggests that industry does not replaces the boilers
homogeneously at the end of a specified technical lifetime, but often they remain in
place after refurbishment also for backup purposes. The industrial sectors use a large
variety of different specific boiler technologies, with different daily load profiles, and
different ways of using them. To capture heterogeneity, the model uses sector-specific
survival functions coupled with economically driven premature replacement or
lifetime extension. In addition, the model uses discrete choice theory formulations for
the mix of fuels and technology types.

The model uses a long list of sectors, subsectors and processes and a specific list of
CHP and boiler plants by sector. This allows capturing CHP dynamics at sector-
specific level and detailed representation of energy savings potentials.

Page 43
 PRIMES 2018

J.4. PRIMES BuilMo: a detailed residential and services sector

model
In the residential sector (and similarly in the services sector), energy is
consumed as input in processes that provide services to the households, such
as space heating, water heating, cooking, cooling, lighting and other needs. The
decision about the level of energy consumption is related to the need for
services covered by energy and depend on efficiency as well as on economics,
notably relative costs and prices. Income drives demand for services from
energy. Electricity demand for electric appliances is modelled separately by
type of appliance and by efficiency class.
As fuel mix in thermal purposes mainly depends on the equipment technology
and the type of dwelling, PRIMES decomposes the dwelling stock in different
categories. The number of dwellings in each category changes over time and by
country depending on demolishment, renovation and new buildings as well as
on relative costs and prices of energy. Income growth combined with socio-
economic factors influences the projection of population by type of dwelling.
The fuel mix possibilities for water heating and cooking depend on the fuel mix
in space heating and the dwelling category. For example using gas in water
heating is more common when gas is also used in space heating. Substitution
possibilities thus depend on the main energy pattern of the dwelling and
influences the hierarchy of fuel choices. These are well captured in the model.
Energy performance is largely depending on the characteristics of the dwelling
(thermal integrity) and the technology of the equipment which uses energy.
Thus, spending money to improve energy efficiency is represented at top level
where it takes the form of investment improving thermal integrity of new and
old dwellings and at the equipment choice level where the model represents a
choice between technologies belonging to different energy efficiency classes.
Both choices involve a capital budgeting decision and the model compares
annualised capital costs using discount rates against variable costs; solutions
with upfront costs and utilisation performance leading to reasonable pay-back
periods are selected. Costs of investment in thermal integrity depends on
technological possibilities but also on a large variety of specific features of the
dwelling under renovation. Energy performance of new technologies and
buildings follows standards and codes and for some of the eco-design
regulations and labelling interventions the model shows the effects by
reducing perceived costs of more efficient technologies. Utility obligations,
white certificates and ESCOs allow reducing discount rates, which can be
captured in values as assumed for model inputs.
The energy
efficiency potential
in houses is great. Long-term responses to policies and energy price movements depend on
Tapping on renewal of dwelling and equipment stock, which is generally a slow process.
potential is
challenging as it The model tracks vintages of technologies in the dynamic simulations, and
depends on takes into account normal scrapping functions by type of equipment and
decisions by
millions of dwelling and endogenously driven premature replacement and renovation
heterogeneous which depends on economics and policy.
consumers.

Page 44
 PRIMES 2018
Energy efficiency progress saves on variable costs and frees up income
resources, which may be consumed back in purchasing various commodities
and in increasing energy demand as well. Therefore, net energy savings are
less than expected due to energy saving upfront expenditures. This is
commonly a rebound effect, which is fully captured in the model as the
formulation solves simultaneously for useful energy demand, fuel mix, and
energy saving investment while being subject to budget constraints.
Short-term responses to policies and energy prices are also possible. They are
modelled as behaviour-driven changes, notably as modifications of the level of
useful energy services. Reducing temperature level of the heating thermostat,
switching-off lighting, reducing stand-by time of appliances etc. are such
behavioural responses, which imply lower demand for energy but also higher
disutility costs that the model explicitly includes in cost accounts.
Not only energy prices but also specific tariff forms are explicit in the model.
For example it is possible to distinguish between average pricing tariffs and
time-of-use tariffs for electricity, the latter having larger effect on demand for
electricity in peak load times.
Useful energy demand depends also on socio-economic factors. The model
includes number of persons per household and income per capita as drivers of
energy requirements in a house and as determinants of diffusion pace of
electric appliances. Saturation effects as well as income related elasticities
depend on these drivers. Useful energy demand also depends on climatic
characteristics, which are represented as uniform by country, since PRIMES
lacks spatial resolution below country levels.
Energy meets fundamental needs of households. In developed economies
income elasticity is expected to be less than one, while substitutions by non-
energy commodities are rather limited. However, in partially heated houses
(which is one of the dwelling categories included in the model) income
elasticity can approach or exceed one, at least over a limited period of time. In
specific uses such as cooling and some of the electric appliances income
elasticity may exceed one. Econometrics are used to estimate such elasticity
value by use and by country in the PRIMES model; the estimated parameters
are used at the upper level of the nesting, which corresponds to useful energy
demand. In developed countries the share of energy in total consumption is
close to saturation (taking account of price variations), a fact that explains the
observed asymmetry in price elasticities with respect to positive or negative
shifts. It should be noted, however, that PRIMES is not solely based on such
overall elasticities but on a structural representation of demand and supply.
Nonetheless, the PRIMES results also show asymmetry of responses for
decreasing or increasing energy costs and prices.
It is important to note that the influence on useful energy demand is via not
only the prices of purchased energy forms but also a cost-price index, which

Page 45
 PRIMES 2018
capture all kinds of costs for energy purposes including annualised costs of
investment in equipment, thermal integrity of buildings and energy savings.
The fuel shares, for each category of end-use in which substitution between
fuels are possible, are represented as fuel choice frequencies (which express
the percentage of households that choose a specific fuel to serve the end-use).
These frequencies can change depending on economics but the flexibility of
change depends on turnover of equipment stock and on the type of dwelling.
The probability that a given appliance (for space heating, water heating and
cooking) is chosen to be installed in a dwelling is calculated as a function of a
total perceived cost and of the maturity of equipment (so that inter-fuel
substitution is constrained) and the possibilities depend on the type of
dwelling. The total perceived cost is a function of capital, maintenance and fuel
(operating) cost of the equipment, as well as of the income of households.
Especially, for cooking and water heating it is assumed that the total perceived
cost also depends on the fuel choice made for space heating following a
hierarchical decision-tree approach. Generally, nested Weibull and nested logit
models are used to model choices. The fuel shares obtained are implemented
for new dwellings and for the installation of new equipment due to normal
replacement. As a result, updated fuel shares by end-use are computed,
concerning both existing and new dwellings.
Specific electricity use is considered as an end-use not allowing substitutions.
Demand depends on the projection of number of appliances by type, which is
driven by socio-economic factors, and on efficiency performance, which
depends on technology choice which is modelled using discrete choice
modelling.
J.4.a. Model database
PRIMES BuilMo includes a very detailed database for buildings which has been
constructed by combining a number of sources – see Box 1. The high resolution
of the model allows to better simulate policies and the way they effect the
building stock, as well as increasing the understanding of how higher energy
efficiency targets can be achieved.
Residential sector
The building stock is split into the following categories:

 Single or multi-storey buildings

 Age of construction: the existing building stock has been divided into nine age
bands covering the period 1920-2015
 Spatial Allocation: three regions have been selected urban, semi-urban and
rural, for each Member State
This implies that there are 54 possible building categories for each of the 28 EU
Member States.

Page 46
 PRIMES 2018
Each building category is characterized by its own heating/cooling characterization
(kWh/m2) which represent the average value of buildings in that category. The
values are calculated based on average engineering characteristics of each building in
each category and are calculated based on bottom-up engineering calculations.

The buildings are further characterized by the equipment they use and the fuel
consumption. The model therefore now has 54 different possible building types for
which energy consumption is calculated according to country characteristics which
include size of dwellings, heating degree days and thermostat settings which are also
differentiated by income class.

The classification thus achieved allows to analyse the effect of policies on the building
stock and the barriers which energy efficiency progress faces: e.g. renovating very old
houses leads to high energy savings however it can be difficult because many of these
houses are historic buildings and therefore renovation can be very expensive;
renovating newer buildings is easier however energy savings are lower. Further
different costs and barriers apply to renovation of single vs. multi-family houses: high
cost for single family, lower costs but difficulties in coordinating multiple
owners/tenants in multi-family houses. Following the logic of PRIMES the model
includes both true costs as well as perceived costs which simulate the barriers faced
by the actors.

Services
Also in the services sector the database has been enhanced; the existing three sectors
have been split into the following eight sub-sectors:

 Trade
o Commercial Buildings
o Warehouses
o Cold Storages
 Market Services
o Private offices and other buildings in market services
o Hotels and Restaurants
 Non Market Services
o Public Offices
o Hospitals and Health Institutions
o Schools an Educational Buildings
Different attributes and characteristics apply for each sub-sector with regard to the
condition of the building stock (age, U-values), internal temperature set-points,

Page 47
 PRIMES 2018
ventilation demands; further the entire building stock has been divided by the type of
ventilation into mechanically and naturally ventilated.

Similarly, to the residential sector the high resolution allows to capture the
specificities of the different sectors including the different operating hours, number of
occupants for the various building types. The higher split can also represent
incentives given to public building renovation more efficiently.

Sources database:

Eurostat Database: Housing Statistics in the European Union:

http://ec.europa.eu/eurostat/statistics-explained/index.php/Housing_statistics
National Statistics Bureaus
The Entranze Project, http://www.entranze.eu/ accessed on 10 April 2017
Inspire Archive, http://inspire.ec.europa.eu/webarchive/index.cfm/pageid/6/list/3.html accessed on 15
December 2016.
BPIE, http://bpie.eu/publications/ accessed on 3 November 2016
The Healthvent Project, http://www.healthvent.byg.dtu.dk/ accessed on 10 April 2017
Europe's Building under the Microscope: A Country-by-Country Review of the Energy
Performance of Buildings., BPIE, 2011

Sources engineering calculations:

CIBSE, CIBSE Guide A: Environemental Design, 2007

EN 13790:2008 Energy performance of buildings - Calculation of energy use for space heating
and cooling, 2008
Guide to the design, installation, testing and maintenance of services supplying water for
domestic use within buildings and their curtilages: BS 8558:2015
2013 ASHRAE Handbook: Fundamentals, ASHRAE, 2013

J.4.b. Equipment and electric appliances

The new model version includes a significant higher amount of options for equipment
(boilers for space and water heating, cookers, air conditioning) and electrical
appliances (TVs, etc.) both for the residential and services. Also the fuel options for
the different options has increased. Different kinds of heat pumps (air to air, water to
air…) for space heating and cooling are included in the database.

The higher amount of detail allows to more accurately reflect the eco-design directive
and Regulations as additional equipment/appliances are now explicitly modelled.

The list of appliances now includes:

 Refrigeration

Page 48
 PRIMES 2018
 Freezing
 Dish Washers
 Washing Machines
 Dryers
 Lighting
 Information and communication
 Entertainment
 Vaccuum Cleaners
 Ironing
 Small Appliances

Figure 3: Correspondence of PRIMES and eco-design categorisation for

appliances

Page 49
J.4.c. Renovation strategies
Sources:

The Entranze Project, http://www.entranze.eu/ accessed on 10 April 2017.

Cost-Effective Climate Protection in the Building Stock of the New EU Member States: Beyond
the EU Energy Performance of Buildings Directive, ECOFYS
Andreas Uihlein, Peter Eder, Towards additional policies to improve the environmental
performance of buildings Part II: Quantitative assessment European Commission Joint Research
Centre Institute for Prospective Technological Studies 2009
Study on the Energy Savings Potentials in EU Member States, Candidate Countries and EEA
Countries Final Report Fraunhofer-Institute for Systems and Innovation Research 2009

Renovation strategies are defined as the subsequential choices of, first, a

renovation type, followed by the possible choices of new equipment for the
different needs (first space heating, then water heating, cooling and cooking).
Renovation types are classified into 8 types and refer to the changes
undertaken in the buildings: R1 is the lightest renovation possible and implies
the replacement of windows only, whereas R8 is the deepest form of
renovation which requires, aside from window replacement, also addition of
thick insulation to the building in order to reduce air permeability levels and
leads to consumption below2 40kWh/m2.
Depending on the renovation type chosen, the probability of changing the
equipment for space heating – before the end of lifetime - changes: if only
windows are replaced synchronous change of the boiler is not deemed any
more or less probable than without renovation, whereas with increasing
renovation depth the probability of changing the equipment, particularly to
heat pumps increases. The choice of the equipment technology for water
heating, cooling and cooking is conditional on the choice of the space heating
equipment.
Finally a combination of 1000s of strategies is obtained which are ranked
according to performance; only feasible options are included, also options
which are deemed highly unlikely are excluded from the possibilities.3 The
exclusion of options is undertaken to obtain a balance between available
strategies and reasonable computational time.
The choice of renovation type and larger equipment with long life time is
intertemporal, not myopic: this implies that the system is optimised over the
entire time period to 2050/2070. Contrarily to the choice of electric appliances
like a TV, renovation choices imply large capital investment and long term
benefits in terms of reduction of fuel expenditures, an intertemporal
optimisation is assumed to be the better modelling option.
Fuel switching is possible when changing equipment or when renovating,
however the more accurate representation of the sector in the new model
accounts for the inertia of the system. Fuel switching often implies costs which
go beyond the “simple” choice of the boiler. Availability of the option is crucial

2
3 E.g. a fuel switch from natural gas to solids is not available, as it is not assumed to be likely under any
scenario condition.
 for example possibility to connect to the natural gas or to the district heating
grid: as the model has no explicit geographic resolution this is approximated
through the knowledge of country specificities and this availability is
considered higher for urban areas than rural.
Hidden costs of fuel switching are also taken into account, for example the
requirements for additional space for storing solids or biomass. Moving away
from district heating is also linked to additional costs. The incorporation of all
the costs for fuel switching make the system rather inert and more stable to
(short term?) fuel price changes, better reflecting reality.
Levels of Details
Renovation
Intensity
R0 No Renovation

R1 Light Renovation (Windows Replacement -U value 2.7 W/m2K)

R2 Light Renovation (Windows Replacement -U value 2.7 W/m2K- and

addition of a 5 cm layer of insulation )

R3 Light Renovation (Windows Replacement -U value 2.7 W/m2K- and

addition of a 5 cm layer of insulation and achievement of air
permeability of 27 m3/m2h)

R4 Medium Renovation (Windows Replacement -U value 1.7 W/m2K- and

addition of a 5 cm layer of insulation and achievement of air
permeability of 27 m3/m2h)
R5 Medium Renovation (Windows Replacement -U value 1.7 W/m2K-
,addition of a 10 cm layer of insulation and achievement of air
permeability rate of to 27 m3/m2h)
R6 Medium Renovation (Windows Replacement -U value 1.7 W/m2K- ,
addition of a 10 cm layer of insulation and achievement of air
permeability rate of to 9 m3/m2h)
R7 Deep Renovation (Windows Replacement -U value 1.7 W/m2K- ,
addition of a 20 cm layer of insulation and achievement of air
permeability rate of to 9 m3/m2h)
R8 Deep Renovation (Windows Replacement -U value 1.0 W/m2K- ,
addition of a 20 cm layer of insulation and achievement of air
permeability rate of to 5 m3/m2h)

J.4.d. The heterogeneous consumer/actor

Income classes are included in the model to simulate the heterogeneity of
actors and the idiosyncratic behaviour of consumers. Instead of having one
single actor the model foresees now a variety of actors each with their own
behavioural characteristics. The behavioural characteristics are simulated by
including differentiated discount rates: each income class has its own specific
discount rate with the highest income class having the lowest discount rate
and the lowest income class having the highest discount rate, representing the
difficulty for such users to apply for financing.-
The population has been split into five income classes; for each Member State
the shares and definitions of the different income classes by MS differ.
This modelling feature allows to better represent the barriers faced for the
renovation within a Member State and among Member States: lower income
classes would benefit most from a reduction of running costs –i.e. from higher
EE in their dwellings,however, they often do not have the financial means to
renovate the buildings as this is a capital intensive investment.
The higher detail of the modelling both in representing the building stock, the
available equipment/appliances and heterogeneity of consumers allows for a
more realistic representation of the barriers faced when implementing policies
aimed at increasing energy efficiency in the residential and services sectors.
J.4.e. Mathematical structure and model concept
Mathematically the model is based on the concept of dynamic discrete choice,
where representative agents decide among a finite set of choices the ones that
are economically more cost efficient. This concept applies to renovation and
equipment. The agents are defined as classes and the available choices are
considered renovation strategies in the model.
A strategy includes the timing and the intensity of the renovation and/or
equipment alternation. The agents have perfect? foresight over the whole
projection period and as a result the decisions made are intertemporal. Each
strategy has a certain economic performance depending on the amount of
upfront investment and the annual spending for the energy bill; a discount rate
specific to the characteristics of the agent –therefore the class- and the
inclusion of possible barriers or hidden costs also influence the computation of
economic performance of each candidate strategy.
The model assumes that within each class the agents who decide about
building strategieshave idiosyncratic behaviour and because they have
heterogeneous preferences do not adopt a single best strategy. Instead a
probability density function is used which assigns probabilities of adoption of
the strategies within the set of the most cost-efficient ones. Possible criteria
used for the evaluation of the strategies are the maximisation of the payback
period or the minimization of the levelized cost of energy.
In particular for the decision on renovation, the cost of each strategy is the
result of a non-linear cost function, which mimics the interaction of the agents
with the market: the high probability of taking up a strategy and the
subsequent uptake of the strategy by many classes or within a class implies
that the strategy becomes more expensive (due to limited availability of labour
and materials for the particular strategy within a particular time span). The
increase in the cost of the strategy makes the specific strategy less popular. As
a result the probability of the uptake of the strategy is modified and another
strategy will also enter the market. Thus a more realistic picture of the uptake
of renovation is obtained.
J.4.f. Policy representation in the new buildings model
The model allows as previously the possibility to include and exclude policies
in order to achieve anything from a baseline to a policy scenario with high
energy efficiency or emissions reduction target.
The new improved model allows for a better and more detailed reflection of
policies.
Overarching policies
Energy Taxation: as in the main PRIMES, taxation is explicit in the
composition of the end user fuel prices; it can be modified as required, by fuel
and by end-user; the standard model includes the obligations stemming from
the current Energy Taxation Directive, as well as the latest taxes from the DG
TAXUD tables
Carbon pricing: carbon pricing as a means to reduce emissions can be
implemented in the model in many different forms: a CO2 tax, inclusion in the
ETS of non-ETS sectors, a carbon value, etc.
Renewable energy directive: the model shows how the overall targets set in
the RES directive are achieved in the H&C sector; incentives for RES put in
place by MS to achieve contributions from the H&C sector to these targets can
be explicitly represented. The inclusion of different types of heat pumps
(ground source and air source), allows for a more precise calculation of RES-
shares according to the latest methodology for calculating RES shares.4
Energy Efficiency policies
EPBD & Building codes at national levels: are now explicitly represented in
the model for new, as well as for renovation, when these are clearly defined;
building specifications follow engineering based calculations for the
determination of energy requirements.
Eco-design and labelling directive: the higher number of equipment and
appliances allows for a more precise representation of the eco-design
regulations; each category of appliances in the model now includes fewer
different types of appliances. Minimum energy performance standards can be
explicitly implemented.
Subsidies/financing: discount rate modification or explicit reduction of costs
for fuels or equipment can be represented explicitly.
EED: the EED includes many aspects which have to be differentiated in order
to correctly reflect them in the model. Some examples include5:
 energy distributors or retail energy sales companies have to achieve 1.5%
energy savings per year through the implementation of energy efficiency
measures: this aspect can be verified through the model by undertaking the
quantification of a baseline scenario (defined e.g. by the exclusion of any
new policies after a specific year), and then comparing it to a scenario
where additional policies are included to verify the energy savings
 the public sector in EU countries should purchase energy efficient
buildings, products and services: this can now explicitly be modelled
 every year, governments in EU countries must carry out energy efficient
renovations on at least 3% (by floor area) of the buildings they own and
occupy: this can now also explicitly be modelled and set as constraint in
the model

4 Share of renewable energy based on Directive 2009/28/EC (SHARES tool)

5 Selected from: http://ec.europa.eu/energy/en/topics/energy-efficiency/energy-efficiency-directive
The energy efficiency value (EEV) still exists as a tool in the modelling,
however it is used as a shadow value (marginal benefit) derived from energy
efficiency gains of an energy efficient constraint (target). Bottom-up defined
policies and EEVs can co-exist in the model. As the model formulates
equilibrium conditions, and not an overall optimization, the shadow value of
energy efficiency acts as a dual variable of a virtual energy efficiency
constraint. By varying the level of the energy efficiency value one can increase
or decrease the stringency of the constraint. In the buildings model, the energy
efficiency value measures EUR/toe saved perceived as an extra benefit by the
decision makers. The value applies to all cost comparisons which lead to the
choice of house renovation options and equipment type options. Consequently
the value increase the competitiveness of efficient options and incite the
decision maker to select them.
Figure 4: Correspondence of PRIMES and eco-design categorisation
for space heating
Figure 5: Correspondence of PRIMES and eco-design categorisation
for water heating and air cooling
 J.7. Mathematical Illustration of Modelling Energy Demand of
Stationary Energy Use Sectors in PRIMES model
J.7.a. First Level: Aggregate demand for energy service be sector
For each sector we first identify the level of demand by a sector h for energy service ES h in a
way to link the energy system to the macro-economic system. The concept of the energy
service is similar to that of useful energy and its measurement is deprived of any effect of
energy or technology efficiency.

The energy service is directly linked to a macro-economic activity or income variable X h . The
relationship between the energy service and the activity variable represent the increase (or
decrease) of the volume of energy service as a function of time or economic development. For
example this relationship may represent improvement of comfort enabled by time or income
growth.
We assume that this relationship is of logistic type (S-curve) by introducing a fixed upper limit
ES h
expressing maximum comfort from an energy service per unit of activity (or income). In
Xh

the case of industry, the logistic curve may be decreasing to express technology trends that
improve the productivity of the energy service. In this case, the limit denotes a lower bound of
productivity improvement.
ES h
 Xh
ES h 
  Tp
Xh


1  exp  0 h   1h    X h , RPh ,    


where  denotes a time index.

If 1h is negative, the logistic process is increasing, expressing an improvement of comfort

from the energy service, while if it is positive, the logistic process is decreasing expressing an
improvement of productivity of energy.
The function  represents the steady-state relationship between the energy service and the
explanatory factors, such as economic activity or income, time and the relative profitability of
the energy service.
Therefore, the function  may be written as:

logh   0 h  1h log X h   2 h log RPh   3h  

where RP stands for relative profitability and  is a time index. The parameters  ,  ,  can
be econometrically estimated or determined as a result of a calibration. The relative
profitability of the energy service may be defined along the concept of opportunity cost and
may be determined as a weighted sum of prices and costs of energy uses and processes
contributing to form the corresponding energy service.

J.7.b. Formulation of allocation decisions of the consumer

To meet the demand for energy service, the decision maker (i.e. the consumer who is
representative of a sector) faces allocation decisions. For example, he must decide about the
mix of processes and technologies.

Penetration of new technologies

Some of these processes or technologies are mature in the market, with known economic and
technical characteristics, some other may be new or anticipated to be present in the future.
 For the latter, two types of costs are identified: the engineering cost, which is the technical cost
for delivering a given technology or process, and the perceived cost for the consumer. For a
mature technology these may be identical, but for an emerging one, with a small market share,
the perceived cost can be much higher. The concept of the perceived cost reflects consideration
about the supply of the technology to the consumer, including costs of maintenance and
operation of the technology choice when this is diverging from the dominant choices in the
market.
Similar technology supply considerations can be introduced in the formulation, to reflect
situations in which a mature technology losing market share beyond some level, enters into a
decline process with perceived costs increasing. This could reflect for example, supply-side
deviations from the efficient scale of technology production.
The question then is, how the model dynamics can accommodate the introduction of a new
technology, or new energy form. A new technology starts from a very high perceived cost.
Exogenous shifts, such as subsidisation of a start-up cost, an accelerated technical progress or
the investment in infrastructure (e.g. for a new fuel) provides an initial push that allows an
increase in the market “acceptability” of the technology. The lower perceived costs will
increase the market share of the technology, thereby further reducing the perceived cost (due
both to technology supply and demand side effects) so that the penetration of the new
technology will accelerate.
The specification can allow several situations such as: more than one technologies prevailing in
the market; a technology entering the market only when reaching a critical competitiveness; or
even simulating avalanche effects in technology penetration..

Diversity of consumer decisions in a sector

Any energy model accepts the notion of the “representative” consumer which is well known in
economic modelling. The decomposition by sector or sub-sector or energy use, aims at defining
decision cases in which the discrepancy of decision conditions becomes smaller, but this
decomposition is limited by available statistics.
Therefore, even within each sector the optimality of the shares of technologies or fuels may
vary across the individual decision-makers operating in that sector. This of course is due to the
varying conditions of energy use for each individual case.
To represent the fact that a number of decision makers with varying conditions operate in a
sector, we must introduce a way of aggregating individual decisions to derive the overall
“average” choice of the representative consumer. This way, even unlikely choices of seemingly
higher costs will have a small share. A number of consumers will choose to adopt them,
because they have welfare or cost gains from deviating from the “norm”, due to their special
conditions.
We propose two alternative formulation to accommodate this aggregation of individual
behaviours:

 We introduce a probability distribution expressing the probability to make a choice that is

different from the average. We interpret the average choice as determined along engineering
cost concepts. An individual decision maker may have an optimal share that differs from the
average, but frequency of such choices decreases as it moves away from the average. For the
representative consumer, this situation leads to a U-shaped average cost curve, having as
minimum the average choice (share). The problem in this formulation, is that the average
engineering shares are exogenously projected to the future and therefore the penetration of
completely new technologies cannot be easily simulated.
 An alternative approach, is to introduce the diversity in the function that aggregates the
choices. While the individual consumers do make discrete choices (the corresponding shares
add to unity) the representative consumer is more likely to choose a mix. This means that at
the level of the representative consumer a concave indifference curve reflects the aggregation
of the individual choices. Therefore the aggregation of the choices is not linear. Following the
Dixit/Stiglitz approach6, we specify a CES-type (constant elasticity of substitution) aggregation
function of the shares for the representative consumer.
In both the economic problem of the representative consumer is to minimise cost of meeting
the total energy service by allocating energy flows to alternative processes or technologies.


min z h   c p ,h Q p ,h  Qbp ,h  Q p ,h
p

subject to

 
f Q p ,h  ES h

aj
i, j , p ,h  Q p ,h  bi

where c p ,h is the unit cost function, that depends on the difference between the choice Q p ,h and
the engineering “optimum” Qb p ,h .

The first of the constraints is the demand constraint, with f being the aggregation function
that can be either linear (first formulation above), or a constant elasticity of substitution (CES)
function (second formulation above). The second set of constraints represents any technical
restrictions on the possible choices.
To introduce the new technology penetration mechanism, we may formulate the perceived unit
Q p ,h
cost c p ,h as a logistic function involving the engineering cost c p ,h , the market share ES h
and an exogenous factor indicating the maturity or acceptability of the technology  p ,h .

 p ,h
c p ,h  c p ,h 
 Q p ,h 
exp a  b  
 ES h 

J.7.c. Second level: Uses or processes

The uses or processes may be organised in the form of a network of flows. For simplicity we
assume here that they are organised as a set per sector.

Each use or process addresses a demand for useful energy. The uses or processes Q p ,h
corresponding to a sector deliver the energy service to that sector. We assume that the
allocation of energy flows to the uses or processes within a sector may involve competition, in
which case the uses/processes are substitutable to each other, at least to a certain degree.
Alternatively, the uses/processes may be complementary to each other, in which case energy
flows depend on technical parameters and involve all uses/processes for a given sector.
At this stage, the objective is to determine the optimal shares of each use or process in a sector,
that minimise total cost of delivering the energy service. Constraints on the mix of
uses/processes are included to reflect varying degrees of competition among the
uses/processes.

6 “Monopolistic competition and optimum product diversity”, Dixit and Stigliz, 1977.
For this allocation decision we propose the first of the two alternative formulation presented in
the previous section. As the alternative uses/processes are known from an engineering point of
view and are complementary at some degree, this formulation is more suitable (the
aggregation function is known). Through the cost function appearing in the objective function,
the formulation introduces the diversity of the decision context. It may also involve the
emergence of new processes or uses (e.g. introduction of natural gas equipped homes) through
the market penetration mechanism.
The representative decision maker (in a sector) minimises total cost of delivering the energy
service:

min z h   CPp,h  Q p,h

p map ( p , h )

The unit cost CPp , h is generally non-linear involving the normative engineering cost CPEng p ,h
the engineering “norm” Qb p ,h , the degree of discrepancy  , the market maturity of technology
 p ,h .


 Q  Qb
  
2

 p ,h p ,h  p ,h
CPp ,h  CPEng p ,h  exp 
2  2
 Q 
  exp a p ,h  bp ,h p ,h ES 
 h

The network equilibrium states that:

(1) ES h   Q p, h   p, h
p map ( p , h )

Bounds can also be incorporated in the above scheme:

Q p, h
(2) lo p, h   up p, h
ES h

The problem (3) to (6) decides the optimal shares of processes and uses.

J.7.d. Third level: Demand for technologies and fuels

At a next level for each process or use, the sector can decide between alternative technologies
and ultimately fuels that can be used for each. This is also an allocation decision problem in
which the sector minimises the cost of serving energy demanded by the process or use and
thereby decides the optimal allocation to technology Qt , p ,h and fuel Q f ,t , p ,h .

As at this stage we seek two represent two phenomena: the fact that the technologies are
strongly competing and only a few of them will ultimately survive in the market; the fact that
the substitutability possibilities between technologies or fuels should reflect a preference
mapping of the decision maker (for example coal cannot be considered as a perfect substitute
of natural gas in space heating). Even if the diversity of situations can explain the existence of a
small share of consumers using for example coal for space heating, at the level of the
representative consumer the relative fuel prices provide insufficient explanation. The
phenomenon can be captured by introducing concave indifference curves that aggregate
technologies and fuels. We therefore adopt the second formulation presented above for this
case.
New technologies, or new energy forms for a given technology can again emerge in the model
dynamics. A new technology starts from a very high perceived cost. An exogenous shift, as
explained above, may provide the initial push to trigger a mechanism that accelerates the
penetration if the technology proves to be competitive.
The sector again minimises total costs of fuels subject to the technical restrictions. The cost
function is non-linear as explained above. A non-linear aggregation function (constant
elasticity of substitution or CES) is used to indifference curves between choices.

min z p ,h   Cf f ,t , p , h
f ,t map ( f ,t , p , h )
 Q f ,t , p ,h  Ct t , p ,h  Q f ,t , p ,h 
Cf f ,t , p, h is the fuel cost and Ct t , p, h is the non-linear cost function of implementing a given
technology.
 t , p , h  Qt , p , h 
 Qt , p ,h  t , p , h  
 t , p ,h
Ct t , p ,h  CPEng t , p ,h    e  Qp ,h 

 Q p ,h   Q 
exp a t , p ,h  bt , p ,h  t , p ,h Q 
 p ,h 

for all t  map( t , p, h) .

 t , p ,h represents the effect of market penetration on the reduction of the perceived cost for a
new technology, or for an existing technology, the increase of the perceived cost when its’
market share declines.

 t , p ,h represents diseconomies of scale as a function of the market share and  t , p ,h is again the
market acceptance or maturity of the technology. CPEng t , p ,h is the engineering cost of the
technology
A CES transformation function links fuels and technologies to meet demand by processes and
uses:


    t1,/p,h  Qt , p ,h   
 1  1

Q p ,h 
t map ( t , p , h )

Qt , p ,h  Q f ,t , p , h
f map ( f ,t , p , h )

Upper and lower bounds for flows can again be included both at the level of technologies and at
the level of fuels.
 Q f ,t , p, h
lo f ,t , p, h   up f ,t , p, h
 Qt , p , h
f map ( f , t , p , h )

 Qt , p , h
f map ( f , t , p , h )
lot , p, h   up t , p, h
Q p, h

Energy savings are one of the possible combination of inputs that can be used for any given
technology. These may include direct energy saving measures, but also other techniques like
for example heat recovery. We introduce a non-linear cost curve with an upper bound which is
the maximum potential that can be achieved for a given process or use.
max
QSav
QSav ,t , p ,h 
,t , p ,h


1  exp a  b  CcSav ,t , p ,h 
solved for the cumulative cost (investment in energy saving) Cc Sav ,t , p ,h which is linked to the
objective function through

Cf Sav ,t , p ,h  CcSav ,t , p ,h .

The demand models compute the demand for each fuel Q f ,t , p ,h and the supply for by-products (as
these have no explicit supply function from the supply side). Some of the fuels (electricity,
steam and probably gas) demanded by the consumers are defined by time segment s . This
notation has been omitted from the above for simplicity.
 K. Transport Energy Demand (PRIMES-TREMOVE)
K.1. Introduction
Energy consumption for transportation purposes generates very significant amount of
greenhouse gases and emission abatement is particularly inelastic in this sector.
Transport is by far the largest consumer of oil products. Expenditures for
transportation purposes represent a significant percentage of GDP.
Because of its importance, PRIMES devotes particular focus on transport and includes
very detailed modelling which covers the energy and mobility nexus and can handle a
large variety of policy measures addressing the transport sector.
PRIMES-TREMOVE Transport sub-model produces projections of transport activity,
stock turnover of transport means, technology choice, energy consumption by fuel
and emissions and other externalities. PRIMES-TREMOVE is a very detailed partial
equilibrium simulation tool used for scenario projections and impact analysis of
policies in the transport sector. The model design focuses on long-term simulation of
conditions, which would drive restructuring of the sector towards new, cleaner and
more efficient transportation technologies and fuels. For this purpose, the transport
model fully handles possible electrification of road transport, high blending of bio-
fuels in all transport sector and market penetration of alternative fuels including
hydrogen. The simulation of dynamics of changes combines modelling of consumer
choices, technology change, refuelling and recharging infrastructure and policy
instruments, which enable the changes.

K.2. Model overview

PRIMES-TREMOVE Transport Model produces projections covering the entire
transport sector by 5-year steps up to 2050. The model projects mobility for
passengers and freight, allocation of mobility by transport mode, projection of
mobility by type of trip, allocation of mobility by mode in transport means,
investment and scrapping of transport means, energy consumption and emissions of
transport means and costs and prices of transport. Choices among alternative options
and investment are specific to each by agent, being a representative of classes of
transport consumers. The choices derive from economics and utility from mobility
and depend on policies, technology availability and infrastructure. The projection
includes details for a large number of transport means technologies and fuels,
including conventional and alternative types, and their penetration in various
transport market segments. The projection also includes details about greenhouse
gas and air pollution emissions, as well as impacts on externalities such as noise and
accidents. Operation costs, investment costs, external costs, tax revenues or subsidy
The transport
sector has a key costs, congestion indirect costs and others are included in the model reports.
role in climate
change strategy Agent choices derive from structural microeconomic optimisation, in which
and for oil technology features and transport activity allocation possibilities are embedded.
independence
policies. Large
restructuring is
required towards
alternative fuels
and technologies.
 Coverage by PRIMES-TREMOVE transport model

Model simulation run: Stand-alone or linked with the entire PRIMES energy system model and the
PRIMES-Biomass model

Time horizon: 2005 to 2050 by 5-year time steps; 2005 and 2010 are calibrated base years and 2015-
2050 being projections.
Countries: Individually all EU 28 Member-States

Transport modes: Private road passenger (cars, powered 2 wheelers), public road passenger (buses
and coaches), road freight (HDVs, LDVs), passenger rail (slow and high-speed trains, metro), freight
rail, passenger aviation (split into distance classes), freight and passenger inland navigation and short
sea shipping, bunkers. Numerous classes of vehicles and transport means with tracking of technology
vintages.

Regions/road types: No spatial resolution below country levels. For trip classes distinction between
Urban areas (distinguished into one metropolitan and other urban areas) and inter-urban areas
(distinguished into motorways and other roads).

Time of day/trip types: off peak and peak time travelling relevant for congestion; passenger trips are
distinguished into non-working, commuting and business trips; freight split into bulk, cargo and
unitized.

Trip distances: stylized histogram of trip types according to distance, representing different agents'
travelling habits per trip and region type.

Energy: all crude oil derived fuels (total and separated by the different grades), biofuels (bioethanol
and biodiesel blends, bio-kerosene, bio-heavy oil and DME), CNG, LNG, LPG, electricity and hydrogen.
Linkage to refueling/recharging infrastructure by trip type.

The exogenous
Emissions: scenario
GHG emissions on assumptions
a TTW and WTW are as follows:
basis, pollutants emissions (CO, NOx, PM, SO2).
Stock of vehicles: Full dynamics of stock turnover for all road (more refined) and non-road transport
 Transport activity for passengers and freight
means.
 Fuel prices, taxation of fuels
 Availability of alternative fuels and regulations on blending
 Other costs and taxations in transport
 Development of refuelling and recharging infrastructure and coverage
 Cost parameters influencing public transport tariffs and infrastructure fees where applicable
 Regulations on technologies, standards on CO2 or on energy efficiency performance of vehicles
 Measures and infrastructure influencing modal shifts and modal efficiency
 Taxations to internalise external costs (for example for air pollution, noise, accidents, etc.)
 Technical improvements and cost changes for various vehicle technologies
 Driving range for battery equipped and hydrogen fuel cell vehicle technologies
 Market coordination assumptions between infrastructure, technology learning and perception of
fuel/technology maturity by consumers.
The model can run as either a stand-alone tool or can run as fully integrated in the
rest of the PRIMES energy systems model. In the integrated run mode, the transport
model takes from the rest of PRIMES projection of prices for fuels, biofuels, electricity
and hydrogen, as well as carbon prices where applicable. The transport model
transmits projection of fuel, electricity and hydrogen consumption to the rest of
PRIMES model. The model linkage supports life cycle analysis of emissions of fuels
used in transport, covering the entire well to wheel calculations. The possibilities,
costs and prices of biofuel supply are assessed using the dedicated PRIMES-Biomass
Supply Model which is also linked with the core PRIMES model and the transport
model, taking from them demand figures and conveying to them bio-energy
commodity prices. Thus, lifecycle analysis of emissions and energy is performed for all
fuel types including alternative fuels.
PRIMES-TREMOVE Transport model can also link with TRANSTOOLS a network
transport model with spatial information. A module handles transformation of
TRANSTOOLS mobility projections in transport activity variables handled by PRIMES.

K.3. Policy analysis focus of PRIMES-TREMOVE

PRIMES-TREMOVE Transport model includes a large variety of policy measures, to
mirror in scenarios. Policy targets, for example on future emissions in transport, are
constraints in scenario projections. The model endogenously determines drivers,
which influence restructuring in transport and substitutions enabling achievement of
the target. The model can handle multiple targets simultaneously. Market penetration
of technologies is not pre-defined but is a result of the model depending on economics
and behaviours. Technology learning is explicit and depends on volume of anticipated
sales.
Market penetration of alternative technologies and fuels in transport heavily depends
on successful market coordination of various agents having different aspirations. At
least four types of agents are identified:

 developers of refuelling/recharging infrastructure aiming at economic

viability of investment depending on future use of infrastructure;

 fuel suppliers who invest upstream in fuel production the economics of which
depend on market volume;

 providers of technologies used in vehicles and transport means who need to

anticipate future market volume to invest in technology improvement and
massive production lines in order to deliver products at lower costs and
higher performance;

 consumers requiring assurance about refuelling/recharging infrastructure

with adequate coverage, and low cost fuels and vehicle technologies in order
to make choices enabling market penetration of alternative fuel/technologies.
The PRIMES model supports explicit analyses of dynamics of market coordination
with individual focus on stylised agents allowing for development of complex
scenarios, which may assume different degrees of success in effective market
coordination. Thus, projections of market penetration of alternative
fuels/technologies are fully transparent and include the entire spectrum of
interactions between consumer choices, technology learning, infrastructure
economics and fuel supply.
The policy measures, which are in the model, can be grouped in soft, economic,
regulatory and infrastructure measures.
 Soft measures include the coordination between the public and the private sector,
information campaigns, certification of services and labelling, partnerships between
the public and the private sector aiming at enhancing knowledge and at using
resources more efficiently. These kind of measures can be mirrored as factors
improving the perceived cost of technologies by consumers, thus allowing for faster
adoption of new or more efficient, but also more expensive, technologies. In the
absence of such measures, the model assumes higher perceived costs in the form of
risk premiums for new technologies, which discourage consumers. Policies that
decrease uncertainty or risk (technical, financing, regulatory. etc.) surrounding
consumer choices can be mirrored by reducing risk premium factors and by lowering
discount rates which are involved in capital budgeting decisions simulated by the
Economic measures model. The perceived cost parameters also reflect anticipation by consumers and can
vary in order to mirror the anticipation confidence by consumers of commercial
maturity of new technologies.
Economic measures aim at influencing consumer choices by modifying relative costs
and prices of fuels and technologies. They include subsidies and taxes on fuels,
vehicles, emissions, congestion and other externalities such as air pollution, accidents
and noise. Certificate systems such as the ETS are also explicit in the model. The level
of the ETS carbon price is determined in the core PRIMES model. Measures supporting
R&D influence costs and performance characteristics of new technologies. Taxation or
subsidisation policies reflect policies at relatively high resolution. They are specific to
individual technologies (e.g. subsidies to BEVs), apply to new versus old vehicles, can
vary by size of vehicle, can link to vehicle performance in terms of efficiency or
emissions, can handle tax exemptions (e.g. exemption from registration tax for new
alternative vehicles), can on fuels, or can vary by vehicle age etc. Economic measures
are also modelled for public transport (for example to influence ticket prices) and for
non-road transport. Fuel taxation is modelled through the standard excise taxes
Regulatory measures which can be defined either in standard form, or in proportion to emissions (direct or
life cycle) or energy efficiency.
Regulatory measures include the setting of targets and technology standards. EU
regulations No 443/2009 and No 510/2011 setting emission performance standards
for new passenger cars and new light commercial vehicles respectively as part of the
European Union's integrated approach to reduce CO2 emissions from light-duty
vehicles are explicit in the model. Tailpipe CO2 emission standards measured in
gCO2/km, which apply on new vehicle registrations are constraints influencing the
consumers' choices upon purchasing new vehicle. In a similar way, energy efficiency
performance standards for all road transport modes have been integrated in the
model; these standards set an efficiency constraint on new vehicle registrations. The
current, as well as future, EURO standards on road transport vehicles are explicitly
implemented and are important for projecting the future volume of air pollutants in
the transport sector and determining the structure of the fleet. The model includes a
special routine, which simulates how the regulations imposing standards influence
supply (structure by technology and vehicle prices) by vehicle manufacturers in order
to influence consumer choices therefore allowing compliance with standards.
Technology standards are also handled in the model for non-road transport
technologies. Targets on emissions or energy can be imposed by transport sector or
overall. Targets influence consumer choices through shadow prices (associated to
Soft measures each target type) which are perceived by the consumers as costs or benefits. Such
Infrastructure policies shadow prices, including carbon values, can be coordinated with the rest of PRIMES
model.
Development of refuelling/recharging infrastructure for alternative fuels (electricity,
hydrogen, LNG, CNG, etc.) is policy driven. Geographic coverage is determined as part
of the policy assumption and concern road and maritime transport. The model
 simulates perception of infrastructure availability by consumers and depending on
the matching between geographic coverage and trip types availability influences
consumer choices. Investment cost recovery options are included. Transport
infrastructure changes and improvements (e.g. intelligent systems, improved
logistics) are not explicitly represented in the model but it is possible in scenario
design to mirror cost, efficiency and modal shift impacts of these policies.

K.4. Novel Model Features

Comparison to The PRIMES-TREMOVE Transport model is a completely new design. It has
TREMOVE model substantially drawn from TREMOVE model but compared to this model the new
design has included significant new developments, which are summarised below:

 Endogenous mileage distribution against various trip types

 Modelling of several additional alternative technologies, fuel types (including

several bio-fuel types) and energy carriers and better representation of
vehicle vintages

 Inclusion of cost-performance (or cost-efficiency) possibility curves for

deriving endogenous technology improvement for conventional and new
technologies in all transport modes

 Detailed modelling of standards on specific CO2 emissions and alternatively on

energy efficiency performance of road vehicles applying on vehicle
manufacturers and influencing supply hence choice of vehicle types

 Integration of multiple parameters in a perceived cost formulation which

captures several factors influencing consumer choice, including "range
anxiety" related to availability of refuelling/recharging infrastructure,
commercial maturity of new technologies and anticipation of policies and
targets

 Formulation of more general discrete choice mathematical functions which

allow representation of consumer heterogeneity through frequency
distributions (histograms)

 Expansion of representation of stylised trip types by type of geographic area

and connection to infrastructure

 Expansion of the modelling of non-road transport, including fast trains, and

maritime transport

 Connection of transport infrastructure development with modal shifts, as well

as with cost and performance characteristics of transport modes

 Endogenous formulation of public transport economics, ticket price derivation

and infrastructure economics including derivation of infrastructure fees

 Lifecycle analysis of energy and emissions by fuel type through linkage with
the entire PRIMES energy systems model
Spatial features The model does not calculate spatial allocation of mobility as the TRANSTOOLS model
which has resolution over a detailed spatial network; spatial coverage in PRIMES-
TREMOVE is stylised and is included for better modelling vehicle choice in relation to
availability of refuelling/recharging infrastructure and thus for treating trip distance
and vehicle ranges as factors influencing choice of vehicle types. PRIMES-TREMOVE
and TRANSTOOLS can interact with each other and exchange data to produce
coordinated scenario projections.
 An important feature implemented in the PRIMES-TREMOVE transport model is the
representation of vehicle range possibilities and the different refuelling infrastructure
development, which influence the choice of vehicle technology by consumers.
Literature indicates that among the barriers for the introduction of alternative fuels
Refueling such as electricity or hydrogen are the "range anxiety" and the lack of
refuelling/recharging infrastructure. Such barriers do not entail direct cost
infrastructure and
implications to the consumers; they rather imply losses to their utility function.
vehicle choice Conventional technologies like ICEs do not neither have range limitations nor face
scarcity of refuelling infrastructure. Vehicles with limited range capability and lack of
refuelling/recharging infrastructure are then endogenously penalised in the model
and thus the corresponding perceived costs by the consumers are increased.
Other barriers are captured through discount rates, which are meant to be subjective
and vary by consumer class to capture different perceptions of opportunity costs of
drawing funds by individuals. Such barriers combined with representation of
Barriers to rational uncertainties surrounding new technologies discourage consumers in opting for
technology choice cleaner and more efficient technologies, which have higher upfront costs and lower
variable running costs. The model can build scenarios in which policies are supposed
to remove such barriers and accelerate market diffusion of cleaner technologies/fuels.
By varying such policies, in intensity and over time, the model can analyse impacts on
diffusion pace and costs arising from eventual lock-ins.
PRIMES-TREMOVE distinguishes a number of different trip types varying according to
purpose, geographic area and time. Average distance by trip type has been estimated
using statistical surveys and depends on area type (metropolitan, motorway, etc.) and
Trip types, “range other factors. Comparing the range possibilities of a vehicle technology against only
anxiety”, and the average trip length of a typical representative consumer is not sufficient to
capture the large variety of situations that exist in reality. Approaches based on
heterogeneity averaging fail to represent the true effects of range limitations on consumer choices.
For this purpose, the model representation of trip categories was extended by
introducing a distribution of trip lengths for each trip category of the model.
Heterogeneity is captured by assuming that a frequency distribution applies on each
trip type showing different frequencies of various trip distances (short, long, etc.). The
distributions have different shapes and standard deviations depending on the trip
nature. By taking into account the distributions, the model compares the range
possibilities of vehicle technology against each class of trip length within a trip
category and derives cost penalties in case of mismatch; an example of a trip
distribution histogram for motorway trips is shown in figure. The cost penalties are
aggregated as weighted sums for each consumer type, depending on the involvement
in the various trip categories and the relative distribution shapes in each category.
The numerical parameters of the model reflect strong aversion for trip cases with high
discrepancy between trip lengths and range possibilities of the
technologies. The purpose of the formulation of heterogeneity in
representation of trips is to assess the mileage performance of
specific vehicle technologies (e.g. BEVs) over a fine resolution of trip
distances. Because vehicles of consumers serve various trip types
and various trip distances, vehicle choice is associated to availability
of refuelling/recharging infrastructure. Range anxiety is modelled
as cost penalising factors, which are endogenously calculated at a
fine resolution level. The model can thus assess cost and technology
diffusion implications of recharging infrastructure development
limited to urban centres versus development with wider coverage.
Lack of adequate refuelling/recharging infrastructure is considered
among the major barriers of large deployment of alternative energy
carriers. Insufficient density of filling stations or public recharging
 plugs prevents consumers from using vehicles in all trip distances and in some
geographic areas, which implies additional costs for the consumer if a vehicle with
such limitations is chosen. PRIMES-TREMOVE model captures this mechanism
through modelling of cost penalties related to infrastructure, which enter the
economic choice modelling of consumers. The aim is to perform cost-benefit analysis
of developing new infrastructure: cost of investment would be compared to benefits
in terms of externalities made possible by wider use of alternative fuels/technologies,
which require the infrastructure. The modelling includes the following:

 The "spatial infrastructure" module features exogenous assumptions as regards

refuelling/recharging infrastructure development and distinguishes between
different areas and trips (urban/inter-urban/short/long distance). Trip types are
represented through frequency distributions, as mentioned above.
 The "infrastructure finance" module performs financial analysis of refuelling/re-
charging infrastructure, evaluates investment and O&M costs of the infrastructure
by category, and determines fees which optionally applies on users of
infrastructure or are socialised..
 The "market" module dynamically estimates the rate of use of infrastructure
simultaneously with projection of market penetration of alternative
fuels/technologies, depending on availability of infrastructure and its cost of use.
The model projects mileage for each road vehicle type and its distribution over trip
types and regions. Mileage estimation is simultaneous with distribution of mobility
across transport modes and is fully embedded in the utility/cost optimisation of
consumer behaviour. Mileage distribution depends on fuel type, vehicle age, variable/
fuel costs, perceived costs and cost penalties related to availability of refuelling
infrastructure, range limitation and uncertainty surrounding new technologies. The
aim is to capture "real life" driving patterns of potential users of new technology
vehicles (e.g. commuting urban trips). The availability of refuelling/recharging
infrastructure implies that the user cannot use his vehicle in all areas but only at those
covered with adequate density of filling stations.

Endogenous vehicle PRIMES-TREMOVE represents a large set of alternative vehicle technologies, including
conventional IC engines with various fuel possibilities, plug-in hybrid electric vehicles
mileage (PHEVs), BEVs and fuel cell electric vehicles (FCEVs). PHEVs and BEVs include
technology variants with various electric ranges depending on battery capacity (e.g.,
PHEVs distinguishes between 20, 40, 80 km electric range categories and types with
range extenders). Flexible fuel vehicles (FFVs) being able to run on high ethanol-
gasoline blends, vehicle types that can use low blends of biofuels (e.g. E10, B20 etc.),
other biofuels such as biogas, bio-kerosene in aviation, bio-heavy oil in inland
navigation are in the list of alternative technologies represented in the model.
Electricity and hydrogen have been included in road transport for all transport means
and LNG for road freight transport and inland navigation.
Vehicle technologies
and vintages
 Additionally, a fuel choice module
has been developed simulating the
choice of the consumer between
different substitutable fuels upon
refuelling the vehicle. For example, a
diesel car is represented as being
able to run either on conventional
diesel (with low bio-fuel blending)
or on various higher blends of
biofuels (e.g. B20, B100). The fuel
choice lies within the context of
minimizing expenses allowing policy
measures to influence the choice
towards cleaner fuels. Not all
technology and fuel options are
available in base year, but are
assumed to become available
gradually over time and reach
commercial maturity at various
degrees and at different future
times, depending on market uptake.
PRIMES-TREMOVE fully keeps track
of technology vintages for transport
means. New vintages incorporate
the latest technologies and have to
meet standards and regulations,
such as the EURO standards. Second
hand cars are included among the
possible choices of consumers; they
are represented to follow previous vintage technologies and their availability and
prices are calibrated to real market characteristics by country. Trade of second hand
cars between countries is not included in the model.

CO2 Car Standards and Aiming at reducing vehicle tailpipe CO2 emissions the regulations No 443/2009 and
No 510/2011 have set emission performance standards for new passenger cars and
Car Efficiency new light commercial vehicles. The standards apply on average sales of car
Standards manufacturers. PRIMES-TREMOVE modelled these standards as a constraint on
weighted average emission performance of new cars in each period simulated by the
model. A CO2 emissions label is associated to each car type, as included in the model.
Using projected new car sales by type as weights, the model calculates average
emission performance of the new car fleet, which is compared against the standard. If
average performance exceeds the standard, a cost penalty applies on car costs
proportionally to the CO2 label for cars with labels exceeding the standard. Thus,
consumers are incited to modify the mix of car types in their choices; cost penalties
increase until the standard is exactly met in each period. The modelling method is
equivalent of assuming that car manufacturers define high car prices to car types with
label exceeding the standard in order to obtain a mix of car sales, which on average
complies with the standard. The car labels defined as specific CO2 emission
performance (in gCO2/km) are based on the NEDC test cycle. In a similar way,
PRIMES-TREMOVE implements energy efficiency standards (with labels expressed in
toe of final energy per vehicle-km) and can also handle efficiency standards based on
primary energy or emission standards (for various pollutants) based on lifecycle
emission calculation. The model can also handle co-existence of multiple car
standards. The same methodology applies also on heavy duty vehicles and other
transport means to capture the effects of new regulations which may apply in the
future.
 Energy efficiency improvement possibilities as an increasing function of unit cost are
represented for all types and technologies of transport means. The cost-efficiency
curves are shown to change over time because of autonomous (market-driven as
opposed to policy-driven) technical progress. Depending on scenario context, the
Efficiency-Cost Curves projected carbon prices or other shadow prices associated to policy targets are
for all transport means modelled as drivers of consumer choices towards more efficient transport means,
which have nonetheless higher unit costs. Therefore, the level of efficiency progress is
endogenous in the model and is derived simultaneously with other variables from
economic optimisation of consumer choices. The inclusion of efficiency-cost
possibility curves is an important mechanism for representing progress of
conventional road vehicle technologies and for capturing efficiency improvement
possibilities for trucks, trains, aircrafts and ships. The model does not include details
about how efficiency improvement is obtained but instead it uses a reduced-form
functional representation of progress enabled by several possible changes, such as
engines that are more advanced, lighter materials, aerodynamic designs, etc. The
numerical estimation of the reduced-form efficiency-cost curves has been based on a
series of engineering studies and laboratory testing reports, which are available in the
literature. The efficiency-cost curves are also fully integrated in the dynamic
representation of technology vintages. For example if assumptions drive early
efficiency progress then future technologies will be at least equally efficient.
PRIMES-TREMOVE transport model is linked with the entire PRIMES energy systems
model and the PRIMES-Biomass Supply model. The linkage calculates lifecycle energy
and emissions of fuels and energy carriers used for transportation.

Lifecycle analysis of The PRIMES projects the entire energy balances and thus calculates primary energy
requirements, which correspond, to the final energy amounts by fuel consumed in
energy and emissions transport. Thus, policy analysis and targets focusing on primary energy or energy
imports can be handled. PRIMES also projects greenhouse gas emissions related to
energy covering the entire chain of energy transformations. Therefore, it can calculate
energy-related lifecycle emissions of transport fuels. Similar lifecycle calculations can
be handled for air pollution. More enhanced air pollution calculations can be carried
out using PRIMES model suite linked with GAINS model (IIASA).
The PRIMES biomass supply model covers the entire lifecycle of bio-fuels and
calculates greenhouse gas and air pollution for the entire chain of transformations,
including cultivation, imports, pre-treatment, transport and conversion of biomass
feedstock into biofuels. So, calculations of sustainability indices can be performed for
all types of fuels used in transport, including mineral oil and bio-fuels (of various
types and based on feedstock of various technology generations).
The entire PRIMES model suite is able to perform calculations of well to tank, well-to-
wheel and tank-to-wheel energy requirements and emissions and to handle policy
targets, standards or taxation associated to such lifecycle indices. The PRIMES suite is
also designed to simulate emission-trading markets (e.g. ETS) which can include parts
or the entire transport sector. Actually, aviation is included in the EU ETS; effects from
that inclusion on costs, prices and efficiency improvement are fully captured in the
model and obviously depend on ETS carbon prices.
 K.5. Demand and supply equilibrium in the transport model
K.5.a. Overview
PRIMES-TREMOVE solves a sort of market equilibrium between demand for transport
Equilibrium between services and supply of transport services.

demand and supply of The model fully captures the features of demand and supply matching which prevail
in transport sector: part of the supply of transport services is carried out by the same
transport services, person who is a demander for such services; in other words, supply is split between
with distinction self-supply of transport services and the purchasing of transport services from
transportation companies.
between self-
There are fundamental differences between self-production of transport services and
production and purchasing from transport businesses: to self-supply the service, the consumer
business production of (individual or firm) faces both capital and variable costs, where capital costs
correspond to the purchasing of transportation means, whereas when purchasing
transport services transport services from transport suppliers the consumer faces only variable costs
(corresponding to ticket prices). Transportation companies also face capital and
variable costs but sell services at transport tariffs (ticket prices, etc.).
In addition, there is no capital rent in self-supply of transport services and the
consumer chooses between alternative self-supply solutions by comparing total costs,
assuming average cost pricing of
Transport alternative solutions. This contrasts
Demand Module prices as set by transportation
companies, which are often based on
marginal costs, which may allow for
Utility/
Max Utility/Min Cost Cost

s.t. G(xi) ≤ a capital rents (e.g. aviation). Other

H(xi) ≥ b Option
1
Option
2
transportation companies owned by the
Option
N
state and subject to strong price
regulation, apply average (instead of
marginal) cost pricing rules to determine
Technology transportation tariffs.
Choice Module
To find the equilibrium between demand
and supply of transport services,
Utility/
Generalised Price Max Profit/Utility
Profit
Demand
s.t. K(xi) ≤ c PRIMES-TREMOVE considers transport
L(xi) ≥ d Option
a
Option
b
Option prices as a pivot influencing both
aN
demand and supply.
To include external costs and also other
Supply costs, such as congestion, the model
includes additional components in the
equilibrium enabling prices which is
termed “generalised price of
transportation” and is calculated both
Equilibrium for self-production and for business
Demand = Supply
supply of transport services.
Based on the above-mentioned
approach, PRIMES-TREMOVE solves an equilibrium problem with equilibrium
constraints (EPEC) simultaneously for multiple transport services and for multiple
agents, some of which are individual consumers and other are firms, which demand
for transport services or produce transport services. The EPEC formulation also
includes overall constraints which represent policy targets (e.g. on emissions, on
energy, etc.) which influence both demand and supply. Mathematically the model
solves as a non-linear mixed complementarity problem.
The transport demand module simulates mobility decisions driven by macroeconomic
drivers, which distribute transport activity over different transport modes and trip
types, to calculate transport services by mode for both individuals and firms. The
decision process is simulated as a utility maximisation problem under budget and
other constraints for individual private passengers and as a cost minimisation
problem for firms.
The transport supply module determines the mix of vehicle technologies (generally
the transportation means), the operation of transport means by trip type and the fuel
mix so as to meet modal transport demand at least cost. In case of supply by
transportation companies, the module calculates transportation tariffs (ticket prices).
Consumer or firm choices at various levels of the supply module use total costs,
inclusive of capital costs, or only variable costs, as appropriate. For example
purchasing a new car involves total cost comparisons among alternative solutions, but
choice of fuel type for an existing car, if that is possible, or determining the rate of use
of an existing car naturally involves only variable costs. The choice of technology is
generally the result of a discrete choice problem, which considers relative costs, which
optionally include factors indicating impacts on externalities.
Solving for equilibrium also includes computation of energy consumption, emissions
of pollutants and externality impacts related to the use of transportation means.
Optionally, policy targets related to externalities (or overall efficiency or overall
emissions) may become binding in equilibrium; through the mixed complementarity
formulation of the model, such overall constraints influence all choices in the demand
and supply transport modules.
Both the demand and supply modules are dynamic over time, simulate capital
turnover with possibility of premature replacement of equipment and keep track of
equipment technology vintages. Foresight assumptions are optional and by default
foresight is limited to two 5-year time periods.

K.5.b. The transport demand module

The transport demand module simulates the decision process of representative
agents in defining total mobility and allocating mobility to a predefined set of
transport modes and of trip types by mode. The model distinctly treats private
passenger transportation and transportation driven by economic activity, such as
movement of products and business trips. The former involves individuals deriving
utility from mobility, whereas the latter involves firms needing mobility for business
purposes.
Representative individuals, i.e. passengers, are formulated to maximise a utility
function subject to income constraint. Utility is derived from transport activity and by
consumption of goods and services not related to transportation. Thus, substitutions
are possible between transportation and non-transportation expenditures, when for
example relative costs of transportation increase. Allocation of income to
expenditures in transportation services and non-transportation goods and services is
derived from optimisation. The projection of income is exogenous and is based on
macroeconomic growth scenarios. Allocation of income to different utility inputs is
organised as a tree involving choices at consecutive levels.
 Utility formation is formulated using a nested Constant Elasticity of Substitution (CES)
function. Concerning transportation-related choices, a first part of the tree involves
trip types, which are organised as a sub-tree, which consecutively deals with trips, by
Business purpose, trips by geographic area and trips by distance classes.
transportation
First sub-tree
includes freight Utility

transport and
Consumption Labour
passenger transport
for business purposes Non-worling
Other non
transport
Other non
transport
trips Commuting
trips

Non
Urban Urban
urban Non urban

Short Long Short Long

Metropolitan Other urban Metropolitan Other urban
distance distance distance distance

Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak
time time time time time time time time time time time time time time time time

Second sub-tree (two examples)

Trip urban

Private Public

Car Non Car Bus Train

Large Small Non

Two wheel Metro Tram
car car engine

Motorcycle Moped

Trip
non-urban

Private Public

Car Non Car Bus Non-road

Large Small Non Motor Other

Motorcycle Train Plane
car car engine way Road

Motor Other Motor Other Motor Other High

way Road way Road Slow
way Road Speed

The second part of the overall tree for passenger transport involves distribution
across transport modes of mobility by trip type. The corresponding sub-tree, allocates
nobility between aggregate transport modes, such as public and private, and further
down it allocates activity to more disaggregated transport modes such as private cars
(disaggregated by size), two wheelers, buses for urban trips, coaches for inter-urban
trips, aviation, rail, inland navigation, metro and trams where applicable.
Activity of business transportation derives from cost minimisation under constraints,
which represent mobility requirements associated to macroeconomic activity, which
is exogenously projected. A nested constant elasticity of substitution (CES) production
function is formulated to simulate substitutions at consecutive levels of a tree
structure. The top level of the tree applies a Leontief decomposition of business
mobility in passenger transportation for business purposes (transport to go to work
places – commuting trips – is included in the transport tree of individuals) and freight
transport. The next levels of the tree decompose mobility by trip type, distinguishing
between geographic area types and trip distance classes. In the following tree levels
decomposition starts from aggregate transport modes (bulk-private and public),
which are further allocated to transport means such as trucks (with size
differentiation), freight trains, maritime, etc.
For both passenger and freight, transport mobility allocation differentiates trips
between trip at peak or at off-peak times. The level structure of the trees allows
specifying different values of elasticity of substitution by level to capture the degree of
substitutability between mobility choices. A low elasticity corresponds to choices that
are close to be complementary to each other (in other words allocation is based on
almost fixed proportions), whereas a high elasticity value signify that choices are
substitutable to each other.
The constant elasticity of substitution functional forms are calibrated to past year
statistics. The official statistics of transport activity (e.g. EUROSTAT and DG MOVE
Pocketbook) include aggregate decomposition of activity. Disaggregation up to the
tree structure of the model has been based on transport surveys, on TREMOVE model
data and on accounting techniques (Excel-based models). Validation of the calibrated
transport demand model has been performed consisting of running the model over a
large set of different assumptions about exogenous parameters, calculating aggregate
elasticities and comparing them to econometrically estimated elasticity values as
reported in the literature.
Generally the values of elasticity substitutions in the CES transport activity functions
are small, which implies that modal shifts are rather inflexible, as confirmed by
several empirical studies found in the literature. Aiming at simulating long-term
structural changes, including in the mix of transport modes, the model includes a
“shifting” technique, which applies on the scale parameters of the CES functions and
allows to represent the effects of policies and infrastructure investments driving
modal shifts at higher degrees than observed in the past. Intelligent transport
systems, new transport infrastructure, congestion management policies acting in
favour pf public transport in the cities, inter-modal facilitation techniques, improved
logistics, etc. are examples of interventions that can accelerate modal shifts, in
particular in favour of public transport and rail. PRIMES-TREMOVE does not
represent these interventions in an explicit manner, because it lacks appropriate
spatial resolution, but it can mirror their effects on modal shifts in scenarios, if
detailed transport studies have measured these effects.
The optimisation models for passenger and for business transport activity uses unit
prices/costs, which are associated to each node of the bottom level of the trees and
refer to specific transport modes for specific trip types. These prices/costs are
calculated in the model of transport services supply. The unit costs of upper tree
levels are calculated from minimum cost functions derived from the optimisation.
 Business transport activity trees (examples)

Cost

Non-freight Freight

Business Other non Other non

trips Transport
tranpsort tranpsort

Non Non urban Urban

Urban
urban

Short Long Short Long

Metropolitan Other urban Metropolitan Other urban
distance distance distance distance

Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak Peak Off peak
time time time time time time time time time time time time time time time time

Freight non
urban

Bulk Public

Truck Cargo Unitised

Public

Small Large Public

Large Truck
trucks IWW Train Public Truck
trucks

Motor Other Motor Other Other

Small Large Motor IWW Train
way Road way Road IWW Train Road
trucks trucks way

Motor Other Motor Other

way Road way Road

Freight urban

Small truck Large truck

 K.5.c. The transport services supply module
The transport services supply module determines the mix of transport means
Overview technologies, the mix of fuels and the rate of use of transport means to meet demand
for transport services as given by the transport demand module. To do this a cost
minimisation model is solved which incorporates discrete choice behavioural models
at various levels. Based on the results of cost minimisation, unit prices/costs are
calculated following explicit pricing and cost accounting rules, which are appropriate
for each transport mode. These unit prices/costs may optionally include external
costs. Demand for transport services depend on these unit prices/costs, and so a loop
is established between demand and supply of transport services.
A specific transport means can serve for more than one trip, which have different
characteristics in terms of geographic area, peak or off-peak time and distance. This is
taken into account in establishing how supply matches demand for transport services.
The unit price/cost by mode depends on the characteristics of the trips served by this
mode.
Stock-flow relationships are fully captured in tracking evolution of transport means
fleet (vehicles, trains, vessels, aircrafts). The model considers stock of transport
means inherited from previous periods, calculates scrapping due to technical lifetime,
evaluates the economics of possible premature scrapping and determines the best
choice of new transport means, which are needed to meet demand. The model also
calculates the degree of using the transport means by trip type and so it calculates the
unit costs of trips. To do this, the fuel mix is also chosen endogenously. The calculation
involves all steps and options simultaneously. Balancing of demand and supply is
obtained for each period. The choices are based on cost minimisation, which include
anticipation factors.
The choices involve adoption of specific technologies and fuel types; technical and
economic characteristics of adopted technologies are inherited in future times when
using the adopted technology. The model follows a vintage capital approach for all
transport means, which means that dynamically it keeps track of technology
characteristics of transport means according to vintages. Not only latest technologies
are available in the choice menu in a given time period; the model allows choice of
older technologies, if that is permitted by legislation, which may have lower costs;
thus the model capture behavioural inertia and also market features, such as the
possibility of purchasing second hand vehicles.
There are several factors influencing the choice of a new transport means. They
Choice of new include payable and non-payable elements. The former include true payments
transport means. (internal costs) and external costs (when internalised); the latter include indirect
costs as perceived by decision makers.
Costing methodology
True payable costs include all cost elements over the lifetime of the candidate
transport means: purchasing cost, which is interpreted as a capital cost; annual fixed
costs for maintenance, insurance and ownership/circulation taxation; variable costs
for fuel consumption depending on trip type and operation conditions; other variable
costs including congestion fees, parking fees and tolled roads.
 To compare candidate transport
Schematic representation of factors influencing choice of new means, a total cost index is calculated
transport means which aggregates all cost elements on
Commercial maturity an annual basis. Only capital costs are
influences technology
Annualised
capital cost upfront costs and so they are
transformed in annuity payments. The
choice Annual transformation uses a discount rate,
Fixed costs maintenance/
insurance
which is conceived as opportunity
cost of drawing funds by the decision
Annual taxes
maker. It is calculated as a weighted
Internal average cost of capital, which adds
costs
equity capital valued at a subjective
Annual fuel
discount rate (which is higher for
costs individuals and lower for business)
Variable
costs
and borrowed capital valued at
Annual costs lending interest rate.
for use of
infrastructure
(tolls) Risk premium is also added which has
GHG several components differentiating
emissions
sectors (private versus public), type of
decision maker (higher risk for
Air pollution
individuals) and type of technology
Decision External (higher risk for yet immature
Accidents
making costs technologies). The capital cost
Noise
parameters can be changed by
scenario and over time so as to mirror
Congestion
policies and evolutions which affect
risk premium factors.
Market The purchasing costs of new
acceptance
of the technologies are assumed to evolve
technology
dynamically, according to learning
curves which depends on cumulative
Density of sales and to technology support policy
necessary
refuelling/
(varying by scenario), reflecting
recharging economies of scale from mass
infrastructure
User production. Similar learning curves
are included for car components such
subjective/
perceived
costs
Range as batteries or fuel cells.
anxiety
(electro-
mobility) Multiple external cost categories are
due to transportation. They refer to
congestion, accidents, noise and air
Hidden costs pollution and they are evaluated in
physical and monetary terms by the
model. Monetary values are based on the Handbook of Internalisation of External
Costs, published by the European Commission.
Other factors, which do not necessarily, imply true payments by the user but may
imply indirect costs are influencing decisions about choice of new vehicles (and
generally transport means). The model includes perceived cost factors reflecting:
technical risk of yet immature technologies, acceptance factors representing market
penetration (this factor serves to simulate accelerated market diffusion), density of
refuelling/recharging infrastructure applicable to technologies using alternative fuels
and those that have range limitations.
 Market acceptance factors are used to simulate circumstances where consumers have
risk avert behaviours regarding new technologies when they are still in early stages of
market deployment. Perception of risk usually concern technical performance,
maintenance costs and operation convenience. When market penetration exceeds a
certain threshold, consumers imitating each other change behaviour and increasingly
accept the innovative technologies giving rise to rapid market diffusion. Both stages of
market deployment are captured in the model through appropriate values of market
acceptance factors, which are part of scenario design. Therefore, the model can
simulate reluctance to adopt new technologies in early stages of diffusion and rapid
market penetration, often leading to market dominance, in later stages.
The decision-making is also influenced by the availability of infrastructure and the
range provided by each vehicle technology; these features are particularly important
when new fuels or new technologies enter the market. In order to represent in a more
Infrastructure refined manner the true effects of the range limitations of some vehicle technologies
influences technology and the lack of adequate infrastructure of alternative fuels, the trip categories
represented into the model are assumed to follow a frequency distribution of trip
choice distances. The model assumes that decision makers compare the range possibilities of
each vehicle technology and the availability of refuelling/recharging infrastructure for
all classes of trip types and trip distances and apply cost penalties in case of
mismatches between range limitations or non-availability of refuelling and trip types
or trip distances. Thus, a vehicle or fuel type may not becoming competitive because
of mismatches compared to other options, which do not present such limitations. The
mismatching considerations do not apply to conventional technologies such as the
ICEs and are relevant for BEVs and FCEVs, as well as for alternative fuels such as
electricity, hydrogen, methane, LNG, biofuels, etc. The refuelling/recharging
infrastructure applies to road and to maritime transport networks and ports,
respectively.
Specific fuel consumption of each vehicle type is endogenously determined by the
model and is calculated based on the COPERT7 methodology. The COPERT
methodology enables calculation of fuel consumption of road vehicles as a function of
Specific fuel their speed, which is determined by the endogenously calculated travelling time and
the average mileage of trips per type of road transport mode. The complete COPERT
consumption methodology has been integrated into the model providing a strong analytical tool for
the calculation of the consumption of various fuels and consequent calculations of
costs. For other technologies not included in COPERT such as BEVs and FCEVs, data
from literature and other studies are used. Similar approaches have been followed in
the model to calculate specific fuel consumption by vehicle type and by trip type for
bus/coaches and for heavy-duty vehicles.
The COPERT methodology enables calculation of fuel consumption of road vehicles as
a function of their speed, which is determined by the endogenously calculated
travelling time, the average mileage of trips per type of road transport mode, the
occupancy factor for passenger trips and the load factor for freight transportations.
The complete COPERT methodology as fully integrated into the model also serves to
calculate emissions of pollutants, including NOx, CO, SO2, PM and VOC.
The calculation of fuel consumption for hybrid vehicles has been modelled in such a
way that takes into account the region in which the vehicle is moving. For urban
regions the fuel savings are significantly higher than in non-urban ones because of
traffic congestion and the slower average speeds that lead to more braking and thus
to more energy regenerated by the hybrid powertrain.

7 COPERT is a software program for calculation of air pollutant emissions from road transport (EEA and
JRC).
Scrapping Fuel
of vehicles
choice As far as plug-in hybrid cars are concerned, they are assumed to operate both as pure
electric vehicles and as hybrids. The electric operation depends on the battery
capacity, which indicates an average pure electric mileage between charges. When the
battery supplies are exhausted, the vehicle switches to a hybrid mode burning
conventional fuel. Plug-in hybrid types with range extending engines are also
included. The model includes pure electric vehicles as following a single all electric
operation equipped with high capacity batteries. Electricity consumption for plug-in
hybrids and pure electric vehicles is being calculated using efficiency figures drawn
from literature.
The choice of technology and fuel type when purchasing a new vehicle is represented
in the model as a discrete choice model following a nested Weibull formulation. The
upper level of the decision tree includes ICE types, battery-based electric cars and fuel
cell cars. The next level distinguishes between conventional, hybrid and plug-in
hybrids. Each of these car types is further disaggregated in technology types,
regarding efficiency for conventional cars, range for electric cars, etc.
Car decision tree

New car
decision

Battery Fuel cell

Gasoline ICE Diesel ICE LPG ICE CNG ICE
electric electric

ord ord Low range

Plug-in Plug-in High range
Conventional Hybrid Conventional Hybrid impr impr EV Medium
hybrid hybrid EV
adv adv range EV

ord ord Low Medium Range ord ord Low Medium Range
impr impr range range extender impr impr range range extender
adv adv adv adv

The model includes possibility of fuel choice for some vehicle technologies. The choice
depends on relative fuel costs of vehicles. Cost penalties apply for fuels with poorly
available refuelling infrastructure. A logistic function is used to calculate the
frequencies of alternative fuel choices. For example, a diesel vehicle can refuel with
diesel blend or pure biodiesel if technically feasible.
The capital vintage model includes normal scrapping and possibility of premature
scrapping for economic reasons.
Normal scrapping is represented using a distribution function (two parameters
Weibull reliability function) with calibrated parameters by country. The distribution
function indicates the survival probability of a vehicle type as a function of time after
date of purchase. The model includes dependence of parameter values on income
expectation, to capture scrapping rates reducing in periods of low economic growth
and increasing in periods of sustained growth. For low-income countries scrapping
rates are high but they may reduce rapidly with economic growth.
 Low usage rates of yet not scrapped old vehicles is
endogenous in the model through the determination of
annual mileage by type of vehicle and by vintage. The
driver in the model is economic cost of using a vehicle;
obviously costs (fuel and environmental) increase with
age and mileage decreases.
Premature scrapping of a vehicle is endogenous and
occurs when fixed and variable operating costs are
higher than total costs (including annuity payment for
capital) of a new vehicle. To capture other drivers,
related to behavioural features, the model uses a logistic
function to calculate the frequency of premature
scrapping.
The model includes several present and future regulations, which influence choice of
vehicle technologies. The EURO standards on pollutant emission performance are
explicitly represented in the model for all types of vehicles. The model relates EURO
standards with vehicle vintages and it specifies that only vehicle types, which are
compliant with the applicable EURO standard, are available for choice in each period.
The standards on specific CO2 emissions (e.g. EU regulations No 443/2009 and No
510/2011) are modelled as constraints applying on average emission performance
over all new vehicles that are available for choice. It is assumed that average specific
CO2 emissions of the fleet sold by manufacturers in a period must not exceed the
specific emission standard as applicable, otherwise a high penalty applies. The specific
Regulations CO2 emissions of each vehicle are measured through the New European Driving Cycle
(NEDC). A CO2 label is thus associated to each vehicle type.
influencing choice of
vehicles Average label for new registrations is computed by weighting labels by vehicle type
using the shares of each vehicle type in new registrations. These shares are
endogenous in the model and depend, among others, on the costs of purchasing new
vehicles. If the average label is higher than the applicable standard, the model applies
a cost penalty on the purchasing costs of each vehicle type proportionally depending
on the difference between the vehicle’s label and the standard. As the purchasing
costs of vehicles are modified, consumers are simulated to change the decisions and
so the mix of new registrations is modified towards a lower average label. This
process continues until average label is exactly equal to the standard.
 Transport
Demand Module

Utility/
Max Utility/Min Cost Cost

s.t. G(xi) ≤ a
H(xi) ≥ b Option Option Option
1 2 N

Technology
Choice Module

Utility/
Profit Demand
Generalised Price Max Profit/Utility
s.t. K(xi) ≤ c
L(xi) ≥ d Option Option Option
a b aN

CO2 Regulation

penalty NO
Comply

Supply YES
Equilibrium

Demand = Supply

The model also represents other labelling policies and standards, as policy options.
Energy efficiency labels and standards is such an example. They can be measured
either in final energy or in primary energy terms. Mixed labelling and standards are
also possible.
Obviously, the choice of standards influence future mix of vehicles and this is fully
captured in the model. For example, very strict end-of-pipe CO2 standards would
equally incite battery-based and fuel cell cars, but strict final energy efficiency
standards would promote battery-based rather than fuel cell cars.
Moderate CO2 or efficiency standards can be met also by conventional car
technologies if they become more efficient. Cost-efficiency curves are modelled for all
conventional technologies (and for various technologies and vehicle types in road
transport) to represent a locus of efficiency improvement possibilities. The cost-
efficiency curves have a time dimension and have increasing slopes, which signify that
purchasing costs increase with efficiency but the incremental costs decrease over
time.

Cost of time influences Cost of time represents a monetary valuation of travelling time, which differs between
individual and business passengers, and also differs among transport modes
transport costs depending upon temporally and geographically features. Cost of time is subdivided
into cost of time for non-road and road transport. Cost of time is expressed as the
product of travelling time and the value of time, used to represent the value of travel
time, which differs between the trip types. Travel time is directly influenced by traffic
congestion and for road transport, a congestion function is used. For public transport,
cost of time also includes waiting time, which is also influenced by congestion.
The travelling time is calculated with distinction between metropolitan, other urban,
motorway and other road areas, and depends on allocation of mobility to different
 trip types as calculated in the transport demand module. Travelling time also depends
on exogenously defined parameters denoting infrastructure investment and
expenditures for the creation of parking places. Travelling time for non-road
transport is exogenously defined, taking into account average mileage and speed.
Cost of time is included in the calculation of generalised price of transportation.
Demand for rail transport (passengers and freight) as well as substitutions between
rail and road transportation are covered in the transport demand module. The
transport supply module aims at finding the mix of train types and fuel types to meet
demand. For this purpose, a discrete choice methodology determines the structure of
the train fleet, by distinguishing between metro, tram, urban and non-urban trains as
Rail transport well as high-speed rail. A capital vintage approach is implemented also for rail. Choice
of new types of rail transport is simulated through a logistic share function that
depends mainly on total operational costs and takes into account capital costs, fuel
consumption, emissions etc. The stock of existing rail infrastructure is taken into
account through an aggregate indicator, which influences the degree of renewal of the
train fleet. The model endogenously calculate mileage per vehicle technology, rail type
and train vintage by taking into account relative variable costs and the influence of
regulations. The model includes engineering-based formulas to calculate specific fuel
consumption by train type and vintage and thus it derives total fuel consumption and
emissions. Cost-efficiency curves, conceived as reduced-form representations of
various efficiency improving techniques, are included for train technologies.
Demand for air transport distinguishes between trip
distance classes and between domestic, intra-EU and
international flights. The air transport supply module Airplane distance classes
determines investment in new aircrafts, finds a mix of
stylised aircraft technologies, and calculates fuel < 500
Air transport consumption and emissions. The model includes a few 500 - 1000
stylised aircraft technologies, namely ordinary,
1000 - 1500
improved and advanced which have in that order have
higher investment costs and higher energy efficiency. 1500 - 2000
The efficiency possibilities draw on aggregate cost- >2000
efficiency curves, which are parameterized based on
literature data. Specific fuel consumption is based on engineering-type formulas,
drawn from literature, and the calculation distinguishes between distance classes of
flights. The only alternative fuel possibility is to use blends with bio-kerosene. The
blending rates are exogenously defined and are depending on emission reduction
objectives (signalled through carbon prices) and assumptions about biofuel supply
possibilities (which are included in the biomass supply and bio-fuel blending models).
Inclusion of aviation in the EU ETS is explicitly modelled.
Maritime transport refers to inland navigation and distinguishes between short sea
shipping and inland water ways, as well as between freight and passenger transport.
Vessel types refer to stylised technologies (ordinary, improved, advanced). Cost-
efficiency curves capture possible energy efficiency improvement is relation to capital
costs. Choice of fuels include conventional mineral oil, blended bio-fuels and LNG.
Maritime transport
A separate model projects activity and energy consumption for international maritime
bunkers. Activity is projected using a simplified world trade model covering EU
import exports with distinction of ships carrying hydrocarbons, bulk cargo and
containers. Separate drivers are considered for each category and for energy bulk
cargo the model links to energy imports-exports of the EU. Allocation to EU ports is
based on exogenous parameters and time trends. Energy consumption is based on
specific fuel consumption functions, which use cost-efficiency curves to summarise
efficiency possibilities. Alternative fuels include bio-fuels and LNG.
K.5.d. Generalised Price of Transportation
As mentioned before, the transport supply module projects the structure of the
vehicle, train, aircraft and vessel fleet together with fuel consumption and emissions.
The calculations are based on simulated decisions, which can be grouped as follows:
• Normal scrapping
• Premature scrapping of old stock of vehicles
• Requirements for new vehicle registrations
• Allocation of new vehicle registrations into different technologies
• Fuel choice
• Annual mileage per vehicle type and vintage, which is further distributed
by trip type.
At this stage fuel consumption and emissions are calculated. Policy driven regulations
and standards influence the simulated choices.
The above-mentioned decisions imply expenditure for purchasing transport means,
for fixed and variable operating costs and for externalities if and where applicable.
The model calculates an indicator of unit cost of transportation by mode and trip type,
inclusive of all cost elements, the cost of time and external costs if applicable.
The unit cost is based on average costs for self-supply of transportation services and
on tariff setting rules for business supplied transportation services. The rules mirror
current practices and regulations concerning ticket and tariff setting by
transportation businesses and generally combine marginal cost and average cost
pricing. For aviation, marginal cost pricing is assumed to prevail. For rail and road
public transport, average cost pricing is assumed with partial recovery of fixed capital
costs, depending on assumptions about subsidies. Fixed cost recovery is distributed
across customer types using a Ramsey-Boiteux methodology.
The calculated unit cost of transportation by mode and trip type is termed
“generalised prices of transportation” and is conveyed to the transport demand
module where it influences demand for transportation services. The interaction
through the generalised prices of transportation ensures equilibrium between
demand and supply of transport services.
The transport demand and the technology choice modules reach an equilibrium
through the generalised price of transportation. The generalised price is determined
once the structure of the vehicle fleet is defined (at minimum cost) by the technology
choice module to meet the projected demand derived from the transport demand
 Area Distance classes
Metropolitan area -
Other urban areas -
Short distance coverage ( < 100 km )
Motorway Medium distance coverage (100-300 km)
Long distance coverage (> 300 km)
Short distance coverage ( < 100 km)
Other roads
Medium distance coverage (>100 km)

Scenario context
definition
(policies, regulations,
techno-economic
assumptions)

Spatial infrastructure
module
(feeds exogenous assumptions
on density and location of
refuelling/recharging
infrastructure)

PRIMES-TREMOVE
Market regulatory module
(sets the regulatory environment (e.g. Scenario Quantification:
self-regulation, regulated monopoly, etc)  Energy demand, emissions,
for different markets (infrastructure/ fuel vehicle fleet mix, activity, costs
supply/ automotive industry), influences
market anticipation for alternative externalities, etc.
technologies/fuels )  Rate of use of infrastructure

Infrastructure finance
module
(performs financial analysis,
assesses investment, estimates
costs, remuneration and rate of
return)

module. The generalised price of transportation differs among the transport modes
and across the various trips and regions. It is also endogenously defined as a result
from an interaction between the demand and the technology choice modules.

K.6. Refuelling/recharging Infrastructure

As mentioned above, the availability of refuelling or recharging infrastructure has an
impact on vehicle and fuel choices. Aiming also at supporting cost-benefit analysis,
PRIMES-TREMOVE includes a block of modules on refuelling/recharging
infrastructure development.
The refuelling/recharging infrastructure is represented for urban, semi-urban and
inter-urban categories per country, as a density of refuelling/recharging points. The
projection of densities is exogenous and is part of scenario design.
The density of infrastructure of different fuels in various areas is connected to agents'
travelling habits (represented through stylised histograms of trip distances). The
combination is modelled as a driver of vehicle/fuel choice.
The fuel types with explicit infrastructure modelling are grid electricity, hydrogen,
CNG gas, LNG gas, LPG, biogas and liquid bio-fuels (when there are separate
dispensers for bio-fuels). Specific infrastructure assumptions are included for larger
and heavier vehicles like HDVs and buses and for vessels (e.g. LNG in ports).
The infrastructure finance module calculates investment and O&M costs of the
infrastructure by category, as well as revenues, which depend on the scenario
specification about infrastructure tariff method, funding and remuneration. Using
model-derived rates of use of the infrastructure, the module calculates infrastructure
remuneration and capital cost recovery in case exogenously assumed tariffs are
applied only to users of infrastructure. Alternatively, if tariffs are socialised (i.e.
applied on all consumers), the module calculate the level of the tariff as required to
recover capital costs.

K.7. Calculation of external costs

The main external costs in transport are congestion, accidents, noise and air pollution.
Physical and monetary valuation are projected by the PRIMES-TREMOVE model.
The external costs of congestion denote the additional social cost incurred to the
other users of the road infrastructure by an additional car. The model captures
congestion impacts as changes from base year values due to vehicle activity
depending on exogenously assumed changes in infrastructure. The calculation has
limitations due to limited spatial coverage (stylised geographical areas) of the model.
The aim of the model is to include a monetary valuation of congestion in the cost of
time indicator, which influences choices in demand and in supply of transport
services.
Similarly, the model includes a simple calculation of impacts on accidents, which is
based on total activity of vehicles and on exogenous time trends. The impacts on noise
are based on exogenous parameters that are differentiated by type of vehicle and
technology.
The model calculates air pollution emissions as a function of fuel consumption,
depending on vehicle and technology types and depending on standards. Diffusion of
pollution is not included.
Monetary valuation of externalities is based on average values drawn from literature
and from the impact assessment handbooks published by the European Commission.
The model includes possibility to internalise externality impacts in various forms,
such as inclusion of specific constraints (e.g. upper limits on physical evaluation of
impacts) or as taxation on fuels or on vehicle types defined so as to reflect impacts on
externalities. Obviously, the internalisation influences vehicle and fuel choices and
affects cost of transportation.

K.8. Measuring disutility costs

The PRIMES-TREMOVE model has a microeconomic foundation and solves a utility
maximisation problem for the individuals. When unit price of transportation increases
for any reason, consumer’s utility (as well as the transportation activity) may
decrease if substitutions are imperfect. Fuel price rises, taxation increase, emission
constraints etc. are among the causes, which drive reduction in transport activity.
In monetary terms, the utility level changes are measures following the income
compensating variation method. This calculates the additional amount of income that
 consumers would require to allow increase in transport activity to compensate for the
loss of utility due to the rise of unit price of transportation.
The disutility costs thus reflect the losses in utility (due to lower transport activity) of
consumers in the context of a counterfactual scenario compared to a baseline
scenario.

K.9. Source of data and calibration to statistics

PRIMES-TREMOVE transport model is calibrated to 2005 and 2010 historical data.
The main data come from statistics on passenger and freight transportation activity as
available in EUROSTAT databases. Energy consumption is calibrated to EUROSTAT
energy balances. Vehicle stock for road transport is calibrated to FLEETS database
and to EUROSTAT. Rail data come from EXTREMIS database. Initial values for
occupancy, load factors and average vehicle annual mileages are derived from TRANS-
TOOLS and TREMOVE databases; these initial values are further modified using
special routines to calibrate to EUROSTAT more aggregated data. Data on vehicle
purchasing costs draw on “Car prices within the European Union” reports. Excise
taxes derive from DG TAXUD excise duty tables. Aviation draws data from
EUROCONTROL databases and maritime from IMO databases and other sources.
The split of transport
activity by transport
mode, by transport
Split of total activity
per transport mode by
Load factors/
occupancy rates
means and the
trips and purpose allocation to trip types
(e.g. urban, business)
Specific energy
is a complex data-
treatment task. We use
TRANS-TOOLS TRANS-TOOLS
consumption per vehicle
technology, age and
EURO standard a special routine that
uses data from
COPERT

EUROSTAT (aggregate
Calibration outcome:
 Adjusted annual mileage per
figures), TRANS-
Stock of vehicles per vehicle type, age, trip and TOOLS (split of
type, technology and
age Calibration
purpose (in vkm)
 Slight adjustments on COPERT activity of each
FLEETS/EUROSTAT
functions to simulate country transport mode by
specific real-life consumption
 Adjusted load factors and stylised area, such as
occupancy rates
metropolitan,
motorway etc., by
Annual mileage per
vehicle type and age in purpose such as
vkm
commuting, non-
Total energy Total activity per
TREMOVE
consumption per transport mode (e.g. working, business, etc.
transport sector (e.g.
road sector)
private cars)
and time such as peak,
EUROSTAT EUROSTAT
off-peak. The splitting
routines also draws
from TREMOVE data.
Load factors for freight transportation and occupancy rates for passenger
transportation are simultaneously derived in the splitting routine using data from
surveys and minimum-maximum limits to capture differences by trip type. The stock
of vehicles provided by the FLEETS database is at high level of disaggregation (vehicle
size, fuel, engine size for cars and motorcycles, vehicle gross weight for trucks, EURO
standard). The specific energy consumption is retrospectively calculated using the
COPERT methodology, which considers average speed of vehicles at same level of
disaggregation as the vehicle stock. To calibrate annual mileage of vehicles at high
resolution of vehicle types and vintages, expert-driven values are used to reflect that
for example older cars are less used than new cars.
 K.10. Classification of transport means
Category Type Technology
Gasoline Pre ECE, ECE, Conventional, Euro I-V
Bio-ethanol Bio-ethanol blend, E85 FFV
Hybrid Gasoline Euro IV-VI
Plug-in hybrid Gasoline Plug-in hybrid technology
Diesel Euro IV-VI
Small cars
Bio-diesel Blended Bio-diesel
(<1.4 l)
Synthetic fuels Synthetic fuels
Hybrid Diesel Euro IV-VI
Plug-in hybrid Diesel Plug-in hybrid technology
Battery electric Battery electric technology
Hydrogen Hydrogen fuel cell
Gasoline Pre ECE, ECE, Conventional, Euro I-V
Bio-ethanol Blended Bio-ethanol, E85 ethanol car
Hybrid Gasoline Euro III-V
Plug-in hybrid Gasoline Plug-in hybrid technology
Diesel Pre ECE, ECE, Conventional, Euro I-V
Bio-diesel Blended Bio-diesel
Medium Cars
Synthetic fuels Synthetic fuels
(1.4 - 2.0 l)
Hybrid Diesel Euro III-V
Plug-in hybrid Diesel Plug-in hybrid technology
Battery electric Battery electric technology
LPG Conventional, Euro I-V
CNG Euro II-V
Hydrogen Hydrogen fuel cell
Gasoline Pre ECE, ECE, Conventional, Euro I-V
Bio-ethanol Blended Bio-ethanol, E85 ethanol car
Hybrid Gasoline Euro III-V
Plug-in hybrid Gasoline Plug-in hybrid technology
Diesel Pre ECE, ECE, Conventional, Euro I-V
Large Cars
Bio-diesel Blended Bio-diesel
(>2.0 l)
Synthetic fuels Synthetic fuels
Hybrid Diesel Euro III-V
Plug-in hybrid Diesel Plug-in hybrid technology
Battery electric Battery electric technology
LPG Conventional, Euro I-V
 Category Type Technology
CNG Euro II-V
Hydrogen Hydrogen fuel cell
2-stroke technology,
Gasoline, biofuels Conventional

Capacity 50-250 cc 4-stroke technology using

Motorcycles
gasoline/biofuels
Capacity 250-750 cc
or electric motors
Capacity 750cc
Moped Conventional,
Conventional, Euro I-V
Mopeds Gasoline, biofuels
Electric mopeds Pure electric technology
Gasoline Conventional, Euro I-V
Hybrid Gasoline LDV gasoline hybrid technology
Plug-in hybrid Gasoline Plug-in hybrid technology
Diesel Conventional, Euro I-V

Light Duty Hybrid Diesel LDV diesel hybrid technology

Vehicles Biofuels Biofuels
(<3.5 ton) LPG LPG
CNG CNG
Synthetic fuels Synthetic fuels
Plug-in hybrid Diesel Plug-in hybrid technology
Battery electric Battery electric technology
Hydrogen Hydrogen fuel cell

Category Type Technology

Capacity 3.5-7.5 ton,
Conventional
Capacity 7.5-16 ton,
Conventional Methane LPG
Diesel trucks
Capacity 16-32 ton, trucks (LNG) trucks
Heavy Duty Conventional
Trucks
Capacity >32 ton,
(> 3.5 ton) Conventional
Capacity 3.5-7.5 ton, Hybrid Truck diesel hybrid technology , biofuels,
Capacity 7.5-16 ton, Hybrid synthetic fuels

Capacity 16-32 ton, Hybrid

Electric trucks, Hydrogen fuel cell trucks
Capacity >32 ton, Hybrid
Busses-Coaches Diesel Conventional, Euro I-V
 Category Type Technology
CNG CNG thermal
LPG LPG
Busses only Hybrid Diesel Hybrid Diesel technology
Battery electric Battery electric technology
Biodiesel Biodiesel technology
Synthetic fuels Synthetic fuels
Hydrogen Hydrogen fuel cell
Metro Metro Type Metro Technology
Tram Tram Type Tram Technology
Locomotive diesel
Locomotive
Locomotive electric
Passenger Train Railcar diesel
Railcar
Railcar electric
High speed train type High speed train technology
Locomotive diesel
Locomotive
Locomotive electric
Freight Train
Railcar diesel
Railcar
Railcar electric
 K.11. Model outputs
The PRIMES-TREMOVE model as the whole PRIMES suite gives standardised outputs
independently if the requirements refer to a baseline, scenario or variant; the set of
information delivered, the excel files delivered, are the same and an overview of the
model outputs may be found below. The projections cover a time horizon up to 2050
by 5-years steps.

Model output Level of detail

Transport activity By transport mode, type of transport means, by
purpose, by agent and by stylised geographic area and
by trip type
Final energy demand By transport mode and vehicle type and by fuel type
Specific energy consumption Efficiency indicators for all transport means
CO2 emissions TTW By transport modes, vehicle type and fuel
CO2 Emissions WTW By transport mode, vehicle type and fuel
Vehicle stock By vehicle type and fuel type, as well as by vintage
New vehicle stock By vehicle type
Refuelling/recharging infrastructure Density by fuel and by geographic area type; linkage to
trip types
Infrastructure costs for charging and Ex-post calculation based on modelling results, related
refuelling to the level of penetration of the different vehicle/fuel
types and analysis of cost recovery
Investment expenditures By transport means, by mode and by agent
Capital costs related to transport By transport means, by mode and by agent in annual
equipment payment terms. In addition, calculation of additional
capital costs for energy and emissions purposes based
on an incremental cost method.
Fixed operation costs By transport means, by mode and by agent
Fuel costs By transport means, by mode and by agent
Excise duty payments By transport means, by mode and by agent
VAT on fuel payments By transport means, by mode and by agent
CO2 tax payments By transport means, by mode and by agent
Ticket prices for public transport By transport mode and finance balances by mode
Registration and circulation tax By transport mode distinguished between household
payments and business expenditures
EU Emission Trading Scheme By transport mode distinguished between household
payments and business expenditures
Variable non-fuel operation costs By transport mode and vehicle type
Disutility costs For passenger and freight transport
Pollutant emissions (CO, NOx, PM2.5, By transport mode and distinction by trip area
SO2) (urban/inter-urban)
External costs (congestion, accident, By mode vehicle type and trip area
air pollution, noise)

The model output is presented in the forms of excel sheets for all the modelling tools.
The PRIMES-TREMOVE model includes two excels files available for each country, as
well as the EU15, NM12 and EU28 aggregates.
K.12. Transport activity modelling using econometrics in v. 6
K.12.a. Transport activity projections

Within the elaboration of the Reference 2015 scenario, a more sophisticated approach
for deriving the transport activity projections by each MS until 2050 compared to the
previous Reference 2013 was developed and this is the methodology for version 6 of
PRIMES. It employs a combined econometric and engineering approach for deriving
transport activity by transport mode. A considerable enhancement in the transport
sector is that it follows the territoriality principle for the heavy-duty trucks activity (in
both the past and the future years), reflecting transportation activity of vehicles
circulating in the territory of the country irrespective of the nationality of the vehicle.

The econometric methodology employs a two-stage error-correction model, which

correlates transport activity with GDP, fuel prices, length of motorways and total
length of railways. Equations (1) and (2) given below show the basic structure of the
two-stage error-correction model.
𝑛

log(𝑡𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡)𝑖,𝑡 = 𝛼0 + ∑ 𝛼𝑖 log(𝑋𝑖𝑡) + 𝑢𝑖,𝑡 (1)

𝑖=1

Δlog(𝑡𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡)𝑖,𝑡 = 𝛽0 + ∑ 𝛽𝑖 𝛥log(𝑋𝑖𝑡 ) + 𝛾𝑢𝑖,𝑡−1 + 𝑣𝑖,𝑡 (2)

𝑖=1

The term (transport) i,t refers to transport activity of country i for year t, the term Xi,t
refers to the respective explanatory variables (e.g. GDP) correlated with each activity
and the terms ui,t, vi,t are the error terms. The coefficients αi, βi are the respective
estimated short and long-term elasticities used in transport activity projections and
the coefficient γ represents the gravitation towards the long-run equilibrium
relationship (1).

The activity projections have been validated using typical indicators such as activity
per capita and, where necessary, they were adjusted to yield realistic values.
Regarding the split of passenger railways into conventional and high-speed rail, we
followed an engineering approach using as an input the expected development of the
high-speed railways network within each MS along the guidelines of the TEN-T core
and comprehensive network.

Regarding, aviation, the new method provides a split into international intra-EU and
international extra-EU aviation. The new econometric methodology treats separately
the two alternative types of trips due to the different dynamics and the expected
increase of the international extra-EU trips to emerging economies (e.g. China). The
activity projections for aviation have been validated against the most recent forecasts
provided by EUROCONTROL.

Sea freight and bunkers activity is correlated with GDP, fuel prices and international
trade. The total trade for each EU country is an additional key driver of maritime
activity. For the bunkers activity projections a panel estimation approach is applied.
The EU-28 countries are being split into 6 regions and as a consequence 6 different
panel estimations are being produced. These estimated coefficients for the countries
that belong to the same region are used for the country-specific projections. The panel
estimation approach is being chosen in this sector, since the bunkers activity in each
EU country is mostly affected by regional (panel specific) instead of country specific
macroeconomic characteristics.
K.12.b. Data update and calibration of the transport sector

The transport sector database has been considerably updated for the purposes of the
Reference 2015 scenario. It includes the most recent very detailed TRACCS database
that provides the most up-to date information regarding the split of the vehicle fleet
for each EU MS. Apart from the update of the vehicle fleet numbers for the past years,
an update on vehicle taxation, maintenance and insurance has been performed.
Various sources were used to update the model database drawing from the TRACCS
and ACEA databases, the MS replies to the questionnaires and other open access
sources.

The database for the techno-economic assumptions has also been updated to reflect
the most recent changes and the expectations of the vehicle manufacturing industry.
The latter refer, in particular, to the expected evolution of the capital costs of
advanced vehicle powertrains such as battery electric vehicles and plug-in hybrids.
The battery costs for battery electric vehicles and plug-in hybrids have been reduced
in the medium and long-term due to the technical progress and steep learning curves
observed in the recent years in the battery manufacturing industry. In addition, the
additional capital costs for improvements in the conventional technologies have been
slightly modified downwards due to the fact that manufacturers tend to absorb
engineering-related costs in the final vehicle price.

During the calibration phase of the model, complex routines calculate transport
related indicators such as the vehicle mileage or the occupancy and load factors such
that the EUROSTAT energy balances and transport activity figures are respected. For
the purposes of the Reference scenario, the calibration routines have been modified to
include additional constraints on the actual activity of heavy-duty trucks in vehicle-
km. The new constraints, even though they increased the computational complexity of
the model, yield more realistic figures regarding the activity of heavy duty trucks.

K.13. PRIMES-Maritime transport model

K.13.a. Introduction
The aim of the PRIMES-Maritime model is to perform long-term energy and emission
projections, until 2050, for each EU MS separately. The coverage of the model includes
the European intra-EU maritime sector as well as the extra-EU maritime shipping.
PRIMES-Maritime focuses only on the EU MS, therefore trade activity between non-EU
MS is not part of the model. Aggregate trade with non-EU countries by non-EU
geographical zones permit modelling of extra-EU flows. The model captures
competition between short-seas shipping and road freight transport. The demand for
maritime services depends on fuel prices and relative costs.
PRIMES-Maritime comprises a demand module projecting maritime activity for each
EU MS by type of cargo and by corresponding partner. Econometrical functions relate
future demand for maritime transport services with economic drivers including GDP,
energy demand (oil, coal, LNG), international fuel prices, and bilateral trade by type of
product.
The supply module simulates a virtual operator controlling the EU fleet, which
performs the requested maritime transport services and allocates the vessels to
activities in the various markets (the EU MS and the extra-EU area) where different
regulatory regimes may apply (e.g. environmental zones). The fleet of vessels
disaggregates into several categories depending on cargo types. PRIMES-Maritime
utilises stock-flow relationship to simulate the evolution of the fleet of vessels
throughout the projection period.
PRIMES-Maritime solves for a balance between demand and supply of maritime
services, with the demand and supply modules interacting dynamically. The allocation
depends on policy measures such as fuel standards or efficiency improvement
regulations. The PRIMES-Maritime model reports both the volume of trade (in tons)
and the maritime transport activity (in tkm) disaggregated by EU MS, by cargo type
and by geographical region. The model also calculates energy consumption by fuel
type and cargo type as well as CO2 and other pollutant emissions. The model projects
investment costs, which mainly includes new vessel purchases and fuel costs.
The PRIMES-Maritime model operates in two modes, namely: (1) the forecasting
mode and (2) the simulation mode. The running modes particularly refer to the
demand module, which determines maritime activity.
When operating in the forecasting mode, PRIMES-Maritime performs a forecast of the
maritime transport activity by EU country following a bottom-up methodology.
Forecasting draws on econometric estimations of trade activity, in tons, between the
EU countries and each aggregate geographical area. Explanatory factors include GDP,
imports and exports of products such as crude oil, dry bulk products, crops, and
others. The evolution of oil prices is an additional critical variable since fuel costs
represent about half of total operating costs8.

8 Ferrari, C., Tei, A. and Parola, F. (2012). Facing the economic crisis by cutting costs: The impact of low-
steaming on container shipping networks. Paper presented at International Association of Maritime
Economists (IAME) Conference, Taipei, Taiwan, 5-8 September 2012
 Exogenous forecasts
Past trends
- GDP
- GDP
- Population
- Population
- International fuel prices
- International fuel prices
- Projections on energy (crude
- Bilateral trade by EU MS with
oil, coal, LNG)
corresponding regions
- Forecasts on crops
consumption

Maritime Transport Demand

Demand by geographical Demand by type of good/

area ship

Short Sea Deep Sea

Shipping Shipping

Maritime Supply module

Operator of the EU
controlled fleet

Aggregate ship types: Exogenous policy

- Tankers variables
- Bulk carriers - Emissions legislation on
- Containers specific markets (e.g. SECA
- General cargo zones)
Provide maritime
services to - Regulatory efficiency
various markets/ improvement mandates (e.g.
zones EEDI, SEEMP) and
autonomous progress
- Vessels productivity
- Fuel prices
New
investments:
Scrapped/
- Fuel mix
Retired
- Efficiency
Improvements

OUTPUT:
- Activity by EU MS and corresponding partner region, split by
ship/cargo type
- Energy consumption, CO2 and other pollutant emissions by
EU MS
- Costs: Investment expenditures on new equipment, fuel cost

Schematic representation of the PRIMES-Maritime model

Geographical zones
The maritime activity derives from multiplying volumes transported by distance
according to origin-destination matrix, with endogenous allocation coefficients, which
North America
link the ports of the trading partner countries or regions. The accounting follows the
South central America territoriality principle. Total maritime transport activity by EU country, in tkm, is the
CIS sum of activities of the specific country.
North Africa
In the simulation mode, PRIMES-Maritime dynamically matches exogenously defined
Sub-Saharan Africa
projections of maritime transport activity by EU country. The projections come either
Middle east from the same model running in forecasting mode or from external sources. Further,
East Asia the PRIMES-Maritime model determines the allocation of import activity among the
South- South east Asia various regions and types of ships. The methodology for allocating the overall
Oceania
transport activity to the various geographical regions and types of ships/ goods
resembles the bottom-up methodology employed in the forecasting mode.

K.13.b. Processing of data input

A very extensive dataset on bilateral trade, available by EUROSTAT, is the starting
point of the model’s database.
The product types are one of the dimensions of the database and a mapping account
for correspondence between various cargo types and the products transported, based
on EUROSTAT. The expanded dataset provides additional information regarding the
types of goods transported between the EU MS and the corresponding partner regions
and is the basis for deriving future forecasts of transport activity by cargo type. The
product types in most countries follow the NST 2007 classification. However, in some
countries (e.g. France, Netherlands) the distinction of product types follows the
NSTR/24 classification due to data limitations.
Types of products: NST 2007 classification
NST
Product types 2007
Products of agriculture, hunting, and forestry; fish and other fishing products 01
Coke and refined petroleum products 07
Secondary raw materials; municipal wastes and other wastes 14
Food products, beverages and tobacco 04
Metal ores and other mining and quarrying products; peat; uranium and thorium 03
Basic metals; fabricated metal products, except machinery and equipment 10
Machinery and equipment n.e.c.; office machinery and computers; electrical machinery and apparatus n.e.c.; radio, television and
communication equipment and apparatus; medical, precision and optical instruments; watches and clocks 11
Unidentifiable goods: goods, which for any reason cannot be identified and therefore cannot be assigned to groups 01-16. 19
Wood and products of wood and cork (except furniture); articles of straw and plaiting materials; pulp, paper and paper products;
printed matter and recorded media 06
Chemicals, chemical products, and man-made fibres; rubber and plastic products ; nuclear fuel 08
Other goods n.e.c. 20
Goods moved in the course of household and office removals; baggage and articles accompanying travellers; motor vehicles being
moved for repair; other non-market goods n.e.c. 17
Transport equipment 12
Coal and lignite; crude petroleum and natural gas 02
Other non-metallic mineral products 09
Equipment and material utilized in the transport of goods 16
Furniture; other manufactured goods n.e.c. 13
Grouped goods: a mixture of types of goods which are transported together 18
Mail, parcels 15
Textiles and textile products; leather and leather products 05
unknown XXX
 Types of products: NSTR/24 classification
Product types NSTR/24
miscellaneous articles G24
metal products G13
transport equipment, machinery, apparatus, engines, whether or not assembled, and parts thereof G20
cereals G01
foodstuffs and animal fodder G06
oil seeds and oleaginous fruits and fats G07
petroleum products G10
potatoes, other fresh or frozen fruit and vegetables G02
wood and cork G04
textiles, textile articles and man-made fibres, other raw animal and vegetable materials G05
crude petroleum G09
iron ore, iron and steel waste and blast furnace dust G11
cement, lime, manufactured building minerals G14
crude and manufactured minerals G15
chemicals other than coal chemicals and tar G18
glass, glassware, ceramic products G22
leather, textile, clothing, other manufactured articles G23
live animals, sugar beet G03
manufactures of metal G21
solid mineral fuels G08
natural and chemical fertilizers G16
coal chemicals, tar G17
paper pulp and waste paper G19
non-ferrous ores and waste G12

The initial detailed dataset from EUROSTAT includes bilateral trade (denoted 𝑡𝑟𝑎𝑑𝑒)
by type of cargo type 𝑘 (measured in tons) between the corresponding countries (𝑖
and 𝑗 denoting an EU MS and a corresponding trade-partner, respectively). However,
the activity indicators (𝐴𝑐𝑡 measured in ton-km) are essential for PRIMES-Maritime to
calculate energy and emissions. Therefore, to obtain the maritime activity in PRIMES-
Maritime, the volume of goods transported from an origin point to a destination point
multiplies the average distance between the two points. However, the only available
data regarding maritime activity per EU MS are available for 2005 and 2010 and split
into international-intra EU and international-extra EU transport activity per EU MS.
National maritime is part of the PRIMES-TREMOVE model.
PRIMES-Maritime adopts the territoriality principle for the allocation of the maritime
activity per EU MS, based on some assumptions. As regards the trade of goods
between EU 28 MS (i.e. international intra-EU maritime), the transport performance
attribution is 50% to the origin country and 50% to the destination country. The same
“50-50” principle allocation also applies to the EFTA countries (i.e. Iceland, Norway,
Liechtenstein and Switzerland) and the candidate countries. As regards the
international extra-EU maritime, where the corresponding partner is outside EU-28
and is not an EFTA or candidate country, the maritime activity is attributed 100% to
the declaring EU MS country.
A calibration procedure of PRIMES-Maritime ensures retrospective simulations in
accordance with activity statistics. The calibration involves assignment of average
distances to match statistics by trading region based on an origin-destination matrix.
For the international extra-EU maritime, the assignment of typical distances uses
geographic information for stylised trips linking the main ports of the EU country and
the ports of major extra-EU countries9 that perform the majority of trade. Knowing
the typical distances, total maritime activity by EU MS disaggregates into international
intra-EU and extra-EU maritime shipping.

9http://www.searates.com/reference/portdistance/

(accessed on 5th March 2015)

The calibration adjusts average distances between each EU MS and the various region
partners to match activity statistics. For validation purposes, the obtained distances
compare to actual distance data obtained either from maps or from other sources (e.g.
EUROSTAT). The validation performs at the level of grouped activities: Ro-Ro mobile
self-propelled units, Ro-Ro mobile non-self-propelled units, liquid bulk goods, large
containers, other cargo not else specified. The splits by region or country are specific
to each cargo type.
Additional calibration procedures further split the maritime activity into Short Sea
Shipping and Deep Sea Shipping.
The international intra-EU maritime activity lies within the category of Short Sea
Shipping. However, a part of international extra-EU maritime transport activity
categorized as Short Sea Shipping refer to the movement of cargo that takes place
along coastlines on the enclosed seas bordering Europe. The calibration procedure
solely focuses on the international extra-EU maritime shipping between the EU MS
and the various worldwide regions.
Measuring the volume of the vessels by EU MS is not straightforward. According to a
study10 by Oxford Economics, there are a number of ways to measure the volume of
the EU fleet. The “EU controlled” fleet includes ships that their operational activity
lies within the EU even though that they might be flagged elsewhere. Secondly,
measuring the fleet by the flag nationality is another principle, which determines
under which state/ country the ship falls under.
Following the “EU controlled” principle of measuring the fleet is likely to align most
closely with the purposes of the PRIMES-Maritime model, which is to simulate the
maritime transport activity for the EU MS. According to the ISL study,11 the major
countries controlling the EU fleet are Greece, Germany, Denmark and Italy. However,
in contrast with the general approach of EU transport modelling, the maritime model
does not assign the fleet of vessels to the various EU MS but assumes an overall fleet
serving the demand for maritime services for the EU MS. This is equivalent to assume
that an operator determines the usage of the EU controlled fleet to serve the maritime
activity of all EU MS. Data regarding the volume of the EU controlled fleet draw from
the Shipping Statistics Yearbook 2013 database (ISL, 2013). The base year regarding
the fleet of vessels is 2010 based on the ISL study, which provides data for 2009 and
2012; the data for 2010 derive from linear interpolation.
The ISL study provides a disaggregation of the vessel fleet into four cargo categories.
The four main categories are the tankers carrying liquid cargo, the bulk carriers
transporting dry bulk cargo, the containers ships and general cargo vessels. Tankers
further split into oil, chemical and liquefied gas tankers depending on the type of the
liquid transported. The general cargo ships also include the Ro-Ro vessels, which
carry roll-on and roll-off cargo. This further disaggregation has been possible to
introduce in PRIMES-Maritime by using data from the study by Oxford Economics on
the EU controlled fleet by type of vessel and by gross tonnage.
The available deadweight tonnage by type of vessel is also an important variable for
the purposes of the modelling, apart from the fleet of vessels. Deadweight tonnage
represents the maximum permitted load of the fleet of vessels and denotes the
maximum carrying capacity of the ships.
The ISL database provides the distribution of the total available fleet of vessels and
the deadweight tonnage by age. The distribution of the EU controlled fleet by age

10 http://www.safety4sea.com/images/media/pdf/Oxford-Economics-ECSA-Report-FINAL.pdf
11 ISL Shipping Statistics Yearbook
shows a differentiation depending on the cargo type of the vessel. Indeed, the age of
the bulk carriers is relatively low, compared to the other vessel types.
The evolution of the future fleet of vessels also depends on the evolution of the fleet
productivity throughout the projection period. Fleet productivity, which denotes the
amount of transport work per deadweight tonne per vessel type, is measured in
tkm/deadweight tonne. Activity by vessel type comes from the calibration of PRIMES-
Maritime to 2010, whereas the values for the available deadweight tonnes come from
the ISL study. The productivity values for the base year (2010) in PRIMES-Maritime
model differentiate according to the various vessel type categories. The average
productivities by type of vessel are relatively lower than the values reported by the
IMO report. This implies that part of the EU controlled fleet and the available DWT, in
reality, provides transportation services outside Europe. Technical features of vessels
The available fleet of vessels has technical features, which refer to average speed of
vessels and to days at sea spent annually for travelling purposes. The assumptions
presented draw from the IMO GHG Study (2014). The data restrict the mileage that
vessels can perform per year for maritime services.
The low sulphur limits, established in the Emission Control Areas (ECA) to minimize
sulphur and other pollutant emissions from ships, would drive a significant
penetration of alternative vessel engine types using LNG. LNG vessel engines are part
of the model among the alternative technological options. Capital costs for LNG
vessels draw from literature, but may reduce in the future depending on the expected
installed engine capacity. The model includes data on capital cost of the various vessel
types disaggregated by size and engine type. The capital costs increase with the size of
the vessel and the power of the engine. The values of capital costs assumed in
PRIMES-Maritime model refer to new-built vessels. Vessel costs related to energy
efficiency progress add on top of the vessel capital costs.

K.13.c. Supply- Fleet module

A virtual supplier controlling the EU fleet of vessels provides the transportation
services to the various markets (e.g. country specific). The equilibrium ensures that
supply balances the demand for transportation services in every period. The available
stock of vessels evolves dynamically through investment and retirement.
PRIMES-Maritime employs a scrappage function to determine the fleet of vessels of a
specific age that are retired. The scrappage function differentiates by type of vessel, as
average lifetime can be different. The volume of scrapped vessels depends on the
vessel type, the age of the vessels and the total available fleet of vessels in the
previous period.
The model takes into consideration the second hand vessels, which allows for a more
realistic simulation of the evolution of the fleet of vessels throughout the projection
period in terms of both the efficiency improvement of an average vessel, as well as the
associated investment costs. Data regarding the second hand prices of vessels and
contracting prices for newly built ships draw from an ISL study. The model projects a
ratio denoting the share of second hand vessels in the EU controlled fleet as a function
of future prices of ships.
Fuel costs derive from multiplying energy consumption by country by energy prices,
which come from the rest of the PRIMES model. The costs related to capital split by
MS by using an indicator reflecting the degree of operation of specific fleet types in
specific EU countries. Hence, if the virtual operator purchases very efficient vessels to
operate mainly on low sulphur expensive marine diesel fuel in ECA zones, then the
purchasing cost correspond to the countries where most of the activity takes place.
The vessel choice module in PRIMES-Maritime is based on discrete choice theory
modelling. A Weibull functional form is used to determine the frequency of choice of a
certain vessel engine. The choice of fuel by vessel type depend on short-term costs
and the technical possibilities of the vessel type. This implies that a vessel equipped
with an Internal Combustion Engine (ICE) will choose the most economic fuel (RFO).
The LNG engine only uses LNG as a fuel. The higher the size of the vessel implies
higher daily expenditures for operational costs; the crew wages alone account for
about half of the total operational costs. However, strong economies of scale prevail
regarding operational costs, especially when accounting the vessel DWT; the same
also applies for fuel costs. Hence, the average unit cost of operation, inclusive of fuel
and operational costs (in Euro/DWT), decreases for high vessel sizes. The operational
costs in PRIMES-Maritime depend on the age of the vessel as insurance and
maintenance costs change with vessel age.
The specific fuel consumption of vessels draws from relevant literature and the
TRACCS database. The calibration adjust these factors to match country statistics on
maritime activity and fuel consumption. A frequency- Origin-Destination matrix built
for 2010, serves the calibration, showing the frequency of movements (routes) that
each vessel type and size cluster undergoes.
Several alternative fuel options are included in the energy model regarding the fuel
mix of the new vessels. The available fuel options are residual fuel oil, low sulphur
marine diesel oil (LSMDO), natural gas and biofuels; biofuels include particularly the
possibility of using biodiesel produced from advanced processing methods and
biomethane. LSMDO, biofuels and natural gas are fuels that comply with the
environmental regulations in ECA zones. According to the EUROSTAT energy
balances, the maritime sector consumed only fossil fuels in 2010. The model can
handle future changes in the fuel mix depending on relative costs and regulation.
The choice of fuel mix for new vessels based on discrete choice theory derives from a
Weibul functional form, which includes relative unit operation costs of alternative
technologies (mainly fuel costs) and annualized fixed costs for purchasing new vessel
types. Market barriers mirroring availability of supply of various fuels or the market
readiness of new vessel technologies also influence the decision. Market barriers have
the form of perceived costs, which do not correspond to actual costs, but influence the
agent’s decision in the discrete choice modelling by increasing the costs of the
relevant choices.
Regulations can influence the fuel mix. The strict sulphur regulations over the ECA
zones (0.1% sulphur content mandate) would pose constraints on the use of high-
sulphur fossil fuels. The model includes several alternatives to comply with the
regulation, including distillates, LNG, biofuels and other gaseous fuels.
 Vessel fuel
choice

Refined
Natural gas
petroleum
Engine
Engine

Residual Marine
fuel oil Diesel oil

Petroleum
biodiesel LNG biomethane
based

Fuel options for new vessels in PRIMES-Maritime

The supply- Fleet module of the PRIMES-Maritime model solves the problem of
allocating in an optimal way the available vessels to the specified routes and
transports the specified quantities between the origin country and the corresponding
partner. The amount of goods to carry between trade partners is an input from the
demand module of the PRIMES-Maritime model. The various vessel categories in
PRIMES-Maritime (i.e. tanker, bulk carriers, containers and general cargo) can
obviously transport only the relevant types of goods; meaning that a container ship
does not transport dry bulk goods.
The quantity of goods that a vessel can carry depends on its load factor and its
available deadweight tonnage, both measured in tons. The different vessel categories
considered in PRIMES-Maritime are distinct cases of deadweight tonnage, which
implies that their cargo capacity can differ substantially.
The allocation of the various vessels to the possible routes solves the problem of cost
minimization under equality and inequality constraints. The objective function does
not include only actual costs (e.g. fuel and operational costs), but also perceived costs
which introduce penalty factors.
The penalty factors influence the possibilities and the pace of compliance with
possible strict regulations in the ECA zones. For example, a vessel utilizing residual
fuel oil sees significantly increased costs if it operates in the ECA zones. Given that, the
optimization problem aims to minimize costs including penalties, the activity of
vessels utilizing non-compliant fuels increases in trips outside these zones. In
addition, an oil-powered vessel has the possibility to switch fuels, subject to technical
constraints, depending on the region that the vessel is travelling.
The PRIMES-Maritime model allocates the available fleet of vessels to the various
trips for the transportation of the different cargo types on an annual basis. Depending
on capacity and mileage constraints, it can only provide a limited number of trips of
specific distance per year. This technical constraint takes into account the distance
between the corresponding partners and the average speed of the vessel. The model
assumes a number of hours per day that the vessel is travelling and the annual days
spent at sea, to account for the annual distance travelled by the vessels.
The model calculates pollutant emissions (NOx, CO, NMVOC, PM10, SOx). Average
values draw from the TRACCS database and the IMO study.
The Energy Efficiency Design Index (EEDI) applies to the new ships and it aims to
promote the use of more energy efficient equipment and engines on the vessels. This
measure applies in a stepwise manner mandating specific energy efficiency
improvements by vessel, which are supposed to tighten every five year. The reduction
rates of the EEDI apply to a “baseline” representative vessel of the ships built between
2000 and 2010. The Ship Energy Efficiency Management Plan (SEEMP) is an
additional policy, which aims to improve the energy efficiency of vessels in a cost-
effective manner. This policy does not only apply to newly built vessels but also to the
existing fleet of vessels when it is economically viable.
The potential efficiency improvement possibilities take the form of cost-efficiency
improvement curves following the logic of the PRIMES-TREMOVE transport model.
This implies that in order to obtain an additional improvement in the average specific
fuel consumption of the “base” vessel, the operator of the fleet needs to invest an
incremental amount of money reflecting engineering costs associated with these
improvements in fuel efficiency. The actual cost- efficiency improvement potential
draw from literature (IMO studies 2009, 2014). Efficiency progress is endogenous,
driven by fuel prices, environmental policies and standards.
 L. Power and Steam Generation and Supply Models
L.1. Overview
PRIMES includes a very detailed model for electricity generation, trade and supply
and for steam generation and distribution. The model is dynamic, solving over
multiple periods (until 2050), multi-country to capture electricity trading in the
European internal market, and market-oriented as it projects electricity tariffs by
sector/country and closes the loop between demand and supply.
The PRIMES power and steam model applies a sophisticated optimisation algorithm
to handle long-term simulation of power system operation, power plant dispatching,
investment in new or refurbished power plants, supply/distribution, trading and
pricing of electricity between countries and towards customers/consumers. Power
market simulation is simultaneous with simulation of steam/heat market so as to
capture trade-offs between cogeneration and boilers, between CHP and pure-electric
plants and between self-production and distribution of steam/heat.
The PRIMES power and steam model is rich in representation of technologies, market
mechanisms and policy instruments:
• The fully endogenous investment and plant operation modelling covers all
known generation technologies (more than 150 distinct technologies) and
very detailed representation of renewable energy sources, including highly
distributed resources. Several electricity storage technologies are endogenous,
including hydro with reservoir, hydro pumping, compressed air storage and
hydrogen-based storage. Several CHP technologies and their technical
operation limits are also included.
• Daily and seasonal variations are captured through hourly modelling of
several typical days for each year. Data for typical days include power load,
wind velocity and solar irradiance. Load demand is bottom-up built from
projection of energy end-uses at detailed level by the PRIMES energy demand
models. Demand side management possibilities are handled at detailed level
based on technical potential and costs. Highly distributed generation at
consumer premises is also included and is taken into account in calculating
transmission/distribution losses and costs.
• Investment decisions distinguish between green-field development,
construction on existing plant sites, refurbishment and extension of lifetime of
plants and the building of auxiliary equipment (such as DENOX,
Power sector desulphurization, CHP, CCS-ready, etc.). Investment decisions also distinguish
decarbonisation is a between utility, industrial and highly distributed scales.
powerful strategy
allowing electricity • The model incorporates in detail feed-in tariff and other supporting schemes
to substitute fossil
fuels in otherwise for renewables and simulates individual investment behaviour in RES
inflexible final following project-financing considerations.
demand sectors
• The PRIMES power model includes reliability and reserve constraints, such
reserve margin constraints to address forced outages of plants or unforeseen
demand increases. The model is deterministic and handles uncertainty of load,
plant availability and intermittent RES by assuming standard deviations,
which influence reserve margin constraints. Ramping up and ramping down
restrictions of plant operation, balancing and reserve requirements for
intermittent renewables and reliability restrictions on flows over
interconnectors are also included.
• Flexibility and reserve to balance intermittency from renewables is ensured
simultaneously by storage (various endogenous techniques), ramping
possibilities of power plants (which influence plant technology mix) and
demand response.
• Regulations such as the large combustion plant directives, the (optional)
emission performance standards, the best available techniques standards, the
(optional) CCS-ready recommendations, the CHP directive and the Emission
Trading Scheme are fully implemented in the model.
• The PRIMES power model represents the entire system of interconnectors in
Europe, as well as possible AC and DC line extensions (including optional
remote connections with offshore wind power in North Sea and with North
Africa and Middle East).
• The model can perform simulation of different market arrangements within
the internal European market, including market coupling, net transfer
capacity restrictions versus load flow based allocation of capacities and
others.
A novel feature in PRIMES power model is the inclusion of non-linear cost-supply
curves for all types of fuels, as well as for renewable power sources, for CCS and for
nuclear plant sites.
Cost-supply curves are numerically estimated functions with increasing slopes
serving to capture take-or-pay contracts for fuels, possible promotion of domestically
produced fuels, fuel supply response (increasing prices) to increased fuel demand by
the power sector, exhaustion of renewable energy potential, difficulties to develop
CO2 storage areas, acceptability and policies regarding nuclear site development, etc.
The non-linear cost-supply curves are fully included in the power investment and
plant operation optimisation.
The PRIMES power and steam model finds technology and fuel mix by minimising
total system costs over a long period of time, assuming perfect foresight (optionally
myopic foresight limited to medium term).
The PRIMES power/steam financial model is a separate module, which determines
electricity and steam tariffs by demand sector (and sub-sector) so as to recover
power/steam costs. For this purpose, the financial model simulates wholesale
markets, bilateral contracting between suppliers and customers and regulated tariff
setting for grid cost recovery. Electricity/steam prices are conveyed to PRIMES energy
demand models, which further recalculate demand and load profiles.
The PRIMES power and steam model reports on projection by country of power plant
capacities, plant operation (gross and net output), fuel consumption, generation from
renewables, grid losses, emissions, investment costs, operation costs and
electricity/steam prices by sector.
 L.2. Mathematical Structure
The optimisation is intertemporal (perfect foresight) and solves simultaneously:
Mathematical
Structure • a unit commitment-dispatching problem
• a capacity expansion problem and
• a DC-linearized optimum power flow problem (over interconnectors).
The optimisation is simultaneous for power, CHP, distributed steam, distributed heat,
district heating and industrial boilers and satisfies synchronised chronological
demand curves of power, steam and heat, which result from the sectoral demand sub-
models. Dynamically the model applies a full scale capital vintage formulation (keeps
track of plant vintages until the end of the projection horizon). All types of investment
in all types of plants including storage are endogenous, as well as their operation and
consumption.
The unknown variables include

 capacity additions by plant type (several types of capacity investment);

 extension of lifetime of plants after refurbishment, investment in auxiliary
equipment;
 generation of electricity (or steam or heat) from plants on an hourly basis;
 consumption of fuels (use of more than one fuels or blending of fuels in each
plant type is permitted under constraints);
 emissions, CO2 transportation; and storage;
 injection or extraction from storage facilities on an hourly basis; and
 investment in storage equipment.
The input (exogenous) parameters include

 electricity demand and elasticities,

 plant fleet as existing in the beginning of projection,
 planned capacity decommissioning,
 known capacities under construction in the beginning of projection,
 grid loss rates,
 ramping possibilities of power plants by technology,
 technical restrictions of CHP plant operation,
 unit costs of investment by technology, unit variable costs, unit fixed costs,
 fuel prices,
 site development costs,
 parameters used in non-linear cost-supply curves,
 taxes and subsidies,
 ETS carbon prices,
 feed-in tariffs and other parameters for representing RES support schemes,
 costs and potential parameters for transportation and storage of captured CO2,
 costs and potential parameters of storage technologies,
 unit costs of investment in grids, unit operation costs of grids,
 parameters expressing policy instruments and restrictions (nuclear, CCS,
environmental, efficiency, CHP, etc.),
 parameters expressing cost of development of smart grids,
 parameters on uncertainty affecting calculation of reserve margins,
 restrictions on use of interconnectors,
 capacities and electrical characteristics of interconnectors,
 reliability parameters on flows over interconnectors
Non-linear relationships regard the cost of access to resources, such as fuels, RES and
plant sites. Such resources are represented as upward sloping cost-supply curves
 linking unit costs to cumulative exploitation. The cost-supply curves are country and
resource specific and change over time in order to reflect changing conditions about
potential and technology.
The financial and pricing model is a recursive model, which includes mixed
complementarity formulations to solve a cost allocation problem.

The equations below aim at presenting the optimization problem solved by PRIMES

𝑴𝒊𝒏 𝓏(𝐺(𝑖,𝑛,𝑓,𝑠,𝑡) , 𝐾(𝑖,𝑛,𝑡) , 𝐹(𝑖,𝑛,𝑓,𝑠,𝑡) , 𝐻(𝑖,𝑠,𝑡) , 𝑆(𝑖,𝑡) ) (1)

𝐺,𝐾,𝑃,𝐹,𝐻,𝑆

Subject to

∑ ∑ 𝑮(𝒊,𝒏,𝒇,𝒔,𝒕) = 𝑪(𝒊,𝒔,𝒕) + ∑{𝑴(𝒊,𝒃) 𝑷(𝒃,𝒔,𝒕) } + 𝑯(𝒊,𝒔,𝒕) ∀ 𝒊, 𝒔, 𝒕 (2)

𝒏 𝒇 𝒃

𝑷(𝒃,𝒔,𝒕) = ∑ {𝒀(𝒃,𝒊) [ ∑ ∑ 𝑮(𝒊,𝒏,𝒇,𝒔,𝒕) − 𝑪(𝒊,𝒔,𝒕) − 𝑯(𝒊,𝒔,𝒕) ]} ∀𝒃, 𝒔, 𝒕 (3)

𝒊 𝒏 𝒇

𝑭(𝒊,𝒏,𝒇,𝒔,𝒕) = 𝒉𝒓(𝒊,𝒏,𝒕) 𝑮(𝒊,𝒏,𝒇,𝒔,𝒕) ∀𝒊, 𝒏, 𝒇, 𝒔, 𝒕 (4)

𝟎 ≤ ∑ 𝑮(𝒊,𝒏,𝒇,𝒔,𝒕) ≤ 𝒖𝒓(𝒊,𝒏,𝒔,𝒕) 𝑲(𝒊,𝒏,𝒕) ∀𝒊, 𝒏, 𝒇, 𝒔, 𝒕 (5)

𝒇

𝑷𝒎𝒊𝒏 𝒎𝒂𝒙
(𝒃,𝒔,𝒕) ≤ 𝑷(𝒃,𝒔,𝒕) ≤ 𝑷(𝒃,𝒔,𝒕) ∀𝒃, 𝒔, 𝒕 (6)

∑ 𝑭(𝒊,𝒏,𝒇,𝒔,𝒕) ≤ 𝑭𝒎𝒂𝒙
(𝒊,𝒇,𝒕) ∀𝒊, 𝒇, 𝒕 (7)
𝒏,𝒔

𝟎 ≤ 𝑯(𝒊,𝒔,𝒕) ≤ 𝑺(𝒊,𝒕) ∀ 𝒊, 𝒔, 𝒕 (8)

𝟎 ≤ ∑{𝑯(𝒊,𝒔,𝒕) : 𝑯(𝒊,𝒔,𝒕) ≥ 𝟎} ≤ 𝑯𝒎𝒂𝒙

(𝒊,𝒕) ∀ 𝒊, 𝒕 (9)
𝒔

∑ 𝐦𝐢𝐧 𝒖𝒓(𝒊,𝒏,𝒔,𝒕) 𝑲(𝒊,𝒏,𝒕) ≥ 𝒓𝒎(𝒊,𝒕) 𝑴𝒂𝒙 {𝑪(𝒊,𝒔,𝒕) + ∑{𝑴(𝒊,𝒃) 𝑷(𝒃,𝒔,𝒕) } + 𝑯(𝒊,𝒔,𝒕) } ∀ 𝒊, 𝒕 (10)
𝒔 𝒔
𝒏 𝒃

∆ ∑ 𝑮(𝒊,𝒏,𝒇,𝒔,𝒕) ≤ 𝒓𝒓(𝒊,𝒏,𝒕) ∑ 𝑮(𝒊,𝒏,𝒇,𝒔,𝒕) ∀𝒊, 𝒏, 𝒔, 𝒕 (11)

𝒔
𝒇 𝒇

∑ 𝒆𝒇(𝒊,𝒏,𝒇,𝒕,𝒋) ∑ 𝑭(𝒊,𝒏,𝒇,𝒔,𝒕) ≤ 𝑬𝒎𝒂𝒙

(𝒊,𝒕,𝒋) ∀ 𝒊, 𝒕, 𝒋 (12)
𝒇,𝒏 𝒔
Power/Steam in a nutshell, focusing on its essence and avoiding entering into details
that are not necessary for the understanding.

The indices 𝑖, 𝑛, 𝑓, 𝑠, 𝑡, 𝑏 and 𝑗 refer, respectively, to countries, power plants,

fuels, load segments (hours), years, interconnectors and emission types.
Unknown variables are the electricity productions 𝐺, the power plant
capacities 𝐾, the inter-country power flows 𝑃, the fuel consumptions 𝐹,
injections or extractions from storage 𝐻 and storage capacities 𝑆. Equation (1)
is the objective function expressing total inter-temporal system costs and
involving linear and non-linear cost function, including for fuels, RES and plant
sites, where applicable. These are denoted by 𝓏( ) in which discount factors
(weighted average cost of capital) are used to transform investment
expenditures in annuity payments for capital.
Equation (2) imposes balancing between demand and production, including
injection and extraction from storage and flows over interconnections. The
matrix 𝑀(𝑖,𝑏) has 1, 0 and -1 elements denoting network topology. Demand
𝐶(𝑖,𝑠,𝑡) is exogenous at this stage, but it depends on electricity prices through
solving the model as a closed loop between demand and supply. Let us note
that in the actual implementation, losses and self-consumption of electricity by
plants are also accounted for, but this is not shown in for the sake of
presentation simplicity. Also for simplicity reasons CHP, steam/heat
production plants and flows over distribution grids are not shown.
The interconnector power flows are computed in equation (3), where 𝑌 is a
matrix of power transfer distribution factors (PTDF), connecting net country
power excess/deficit with inter-country power flows.
Parameter ℎ𝑟(𝑖,𝑛,𝑡) denote power plants’ heat rate values, and so equation (4)
defines the amount of fuels consumed. Fuel blending is not shown, for
simplicity. This equation has the form of inequality constraint for intermittent
renewables, with ℎ𝑟(𝑖,𝑛,𝑡) denoting hourly availability of renewable resources.

Equations (5) are the power plant capacity constraints, which use planned and
forced outage rates as known parameters. For simplicity, old capacities,
decommissioning, extension of lifetime with refurbishment and distinction of
new investment cases (on site, new sites, auxiliary equipment, etc.) are not
shown.
Equation (6) imposes upper and lower bounds on flows over interconnections
to represent technical, reliability and regulatory limitations. Equation (7)
makes sure that fuel consumption does not exceed maximum available fuel
amounts, if applicable. Equations (8) and (9) ensure that injection or
extraction from storage do not exceed injection or extraction capacity and that
the amount stored in a year does not exceed storage capacity. For simplicity,
losses or electricity consumption (for example in electrolysis to produce
hydrogen) are not shown.
Equations (10) and (11) illustrate reserve margin and flexibility constraints
respectively, where 𝑟𝑚 and 𝑟𝑟 denote system and plant-based technical
parameters. Equation (12) is an example of upper bound on emissions (𝑒𝑚
being emission factors). Similarly RES obligation or security of supply
constraints can be added.
Demand responses are approximated by expressing demand 𝐶 as a function of
marginal prices, which are dual to the balancing constraint. For this purpose
the model is solved as a mixed complementarity problem (Kuhn-Tucker
conditions).
 The model formulates competition between electricity only plants, CHP, industrial
boilers, district heating and between self-supply and grid supply. A stylized graph of
electricity and steam/heat supply is modelled.
Input fuels and energy forms

Utility Power plants Industrial Power plants Industrial Boilers District Heating Plants Highly distributed plants

ELEC CHP ELEC CHP ELEC CHP

Steam Heat Self-supply of

Electricity Self-supply of Self-supply of
Distribution Distribution electricity
Transmission electricity (industry) Steam
Grid Grid and/or heat

High
Voltage
grid
Medium
Voltage
grid
Low
Voltage
grid

Note Electricity Electricity Electricity Steam Heat

Demand High Demand Medium Demand Low Demand Demand
Grid electricity
Voltage Voltage Voltage
Self-supplied electricity
Steam
Energy Demand Sectors
Heat

The above graph is modelled as follows:

a) arrows represent flows of energy which are among the unknown model
variables;
b) boxes represent equipment with given capacities, for which investment,
decommissioning and tracking of technology vintages is endogenous;
c) demand choices between self-supply and grid-supply are endogenous based
on economics and subject to upper and lower bounds expressing technical
possibilities;
d) demand choices drive supply and determine flows over the grids, hence losses
of energy (losses decrease if self-supply increases);
e) CHP plants can jointly produce electricity and steam or heat based on
technical possibilities represented in the model;
f) overall cost optimisation, subject to min and max bounds, drives competition
between CHP and boilers (it is also influenced by policies).
 L.3. Model Features
L.3.a. Representation of Plants
The PRIMES database includes an inventory of all power plants in Europe. The
individual plants are grouped in a large number of categories, defined according to
fuel type, technology, CHP and scale (utility, industrial). A calibration model
determines heat rates and other operational parameters by group of plants, by
simulating the structure of power generation and consumption of fuels in past years
(2000-2010) aiming at matching aggregate figures with published statistics by
Representation of Eurostat. So the PRIMES model data for old plants are close to reality and thus the
Blending of fuels model is validated.
Plants
A power plant category in PRIMES has thermal (fuel consumption function and fuel
blending possibilities), operational (gross and net capacity, ramping rates, technical
minimum, planned and forced outage rates), age (commissioning date, refurbishment
dates, date of planned decommissioning), cost (capital cost, fixed O&M cost, variable
operating cost) and environmental (emissions, pollution prevention equipment)
characteristics. These characteristics are known for plants existing in the beginning of
projections and also for plants which are under construction and their commissioning
date is known. For new plants, which are candidate for investment, the data for plant
characteristics are drawn from a technology forecasting database and are subject to
change over time reflecting technical progress. In future times, each plant preserves
the characteristics as at construction date.

L.3.b. Investment modelling

Investment in new power plants distinguishes between two cases: a) construction on
a new site, b) construction on an existing plant site if available area allows for.
Obviously, the former is more costly than the latter. PRIMES maintains an inventory of
available plant sites, by type of technology, and projects dynamically site availability
depending on projected decommissioning and new constructions.
Optionally the model treats power plant investment in new plants as integer multiples
Investment of generic plant sizes (which are different per plant type).
modelling
The building of equipment in the electricity and steam system requires several years.
This has important implications for planning and plant type choice. The model
considers the financial costs associated to the construction period but ignores the fact
that the plant types differ in construction time, which may influence plant selection in
particularly uncertain circumstances. Furthermore, under the myopic anticipation
regime, the model considers that the plants can be constructed and immediately used
within the 5-years runtime period of the model. In this sense, the model operates as if
the current 5-years period is perfectly known by the decision-maker.
Old plants, as well as new plants, which are projected by the model, are considered for
possible refurbishment when reaching their planned decommissioning date.
Refurbishment implies improvement of technical characteristics and extension of
lifetime, and may include environmental upgrading. Capital costs of refurbishment are
lower than cost of new investment, but the lifetime extension is limited. Exogenous
parameters reflect technical possibilities, or impossibility, of refurbishment for
certain plant categories (defined by country). The refurbishment decisions are
simultaneous with new investment decisions and are based on economics. Fixed O&M
costs are assumed to increase with plant age and this may drive premature
decommissioning under certain economic circumstances.

L.3.c. Blending of fuels

The model also considers the possibility of building auxiliary equipment to an existing
plant or to a new plan later than construction. Auxiliary equipment include pollution
prevention facilities (for NOx, SO2 and/or for PM), CO2 capture equipment and steam
 cogeneration plant extension. Regulations, emission reduction policies, carbon prices
and steam demand are drivers of such plant upgrades. Building auxiliary equipment
implies additional capital costs and may imply reduction of heat rates of plants.
CCS Exogenous parameters are used to represent technical possibilities and restrictions
about building auxiliary equipment. The model also represents plant categories,
which include such auxiliary equipment at construction time. These plant types
(combustion plants with full pollution prevention equipment, CHP plant with several
CHP technologies and CCS plants with several technologies) are more efficient and
less costly than plants for which auxiliary equipment is built at a later stage. Under
good foresight conditions, it is usual that plants with auxiliary equipment at built
times are preferred for economic reasons. However, in other policy circumstances,
e.g., regulations such as mandatory CCS or large combustion plant directives, it may be
economic to retrofit old plants by adding auxiliary equipment.
The model includes an endogenous possibility of using more than one fuel type in
some plant types (both in power plants and in industrial or district heating boilers).
The fuel mix decision is based on economics but the fuel mix are restricted through
exogenous upper and lower bounds. Fuel blending is used in case of co-firing biomass
products, or industrial by-products (e.g. derived gas) and is possible in few
combustion technologies.

L.3.d. Scale of generation activity

The model put emphasis on the representation of plant efficiency, cost and
performance as a function of plant typical sizes. The size classes are associated to
three stylised scales of generation business, namely utility, industrial and highly
distributed scales. Utilities can invest in large size plants and benefit from economies
of scale, while industrial power and steam producers can invest only in relatively
small size plants have better access to steam uses and markets. Distributed generation
plants are very small scale and are very costly but save over grid losses and costs. The
cost gap between plant sizes is considered to decrease over time and thus distributed
generation gradually becomes more attractive. A similar trend has been observed for
combined cycle plants compared to large-scale open cycle plants. Nuclear costs are
also supposed to decrease with scale.
Scale of
generation L.3.e. CHP
Cogeneration is represented as an efficient frontier of possible electricity and steam
activity combinations from a plant. The possibility frontiers are specified for various CHP.
Both the choice of CHP technologies and the operation mode (mix between electricity
and steam production) are endogenous in the model. The operation possibilities are
restricted by feasible combination of electricity and heat output, which are different,
by plant technology. The CHP operational constraints delimit maximum electric
power and minimum steam combinations as a locus of an iso-fuel line.

L.3.f. CCS
The PRIMES database includes three generic carbon capture technologies (CCS),
namely capture at post-combustion stage, capture at pre-combustion stage and the
CHP oxyfuel technology (which consists in burning with oxygen instead of air). The generic
carbon technologies apply to a series of power plant technologies, including
conventional steam turbine plants, supercritical steam turbine plants, fluidized bed
combustion plants, integrated gasification combined cycle and gas turbine combined
cycle plants. The model represents transport and storage of CO2 through reduced-
form inter-temporal cost-supply curves, which are specified per country.
 L.3.g. Nonlinear cost curves
The power agent’s cost
Price or Decreasing
optimization takes into account Unit cost Increasing Returns
non-linear cost supply curves of With scale
Returns
resources used in power With scale
generation, as for example for
fuels, renewables, sites for
nuclear investment, cost of
storage of CO2 captured, etc.
These cost supply curves
incorporate information about
the maximum potential of the
Resource quantity
resources or the decreasing
Nonlinear cost returns of scale associated to the amount used.
curves
L.3.h. Plant dispatching and system operation
The power model can handle preference or obligations about using domestically
produced fossil fuels (e.g. lignite) by assigning low cost values for some of the first
steps of the cost-supply curves for such fuels. In the same way, the model can handle
take-or-pay obligations, for example for imported natural gas. The low cost values
would reflect a virtual “subsidy” on fuel purchase costs, which however is not
accounted for in transactions but only influence economics of fuel mix. Costs used for
price determination and costs reported by the model account for actual payments
(fuel purchase costs, or fuel production costs). Low prices are attributed to by-
products, such as blast furnace gas, coke-oven gas, refinery gas, reflecting only
variable costs
The model represents demand variability for electricity and steam/heat by including
hourly fluctuation of load in typical days (one for winter and one for summer in the
Plant dispatching reduced model version and nine typical days in the extended model version). Data for
and system annual load curves, from the national TSOs and other sources, are aggregated to
obtain equivalent load curves by typical day.
operation
The operation of all modelled plants and the use of energy input resources are
calculated on an hourly basis for each typical day (load segments). Hourly profiles of
intermittent renewable sources are supposed to be known at country level. Load
segment synchronisation also applies for electricity and steam/heat; this feature is
important for capturing the operation of CHP plants and competition between
cogeneration, boilers and distribution of steam/heat.
The model associates a demand fluctuating profile to every use of electricity and
steam or heat included in the demand sector models and for energy demand in the
energy branch. Special focus has been devoted to represent various optional profiles
of recharging car batteries. By adding up the sectoral load profiles, the model
determines aggregate load profiles by country.
Load profiles change over time. By scenario they also vary, depending on the relative
shares of various energy uses, the prices (which are higher for sectors with low load
factors), the degree of energy savings (and the use of more efficient equipment) and
special demand side management measures including smart metering. The latter
motivate battery recharging at off peak hours.
When load profiles become smoother, capital-intensive power technologies are
favoured and reserve power requirements are lower, implying lower overall costs.
The various power plants contribute to reserve power through differentiated
estimates of their capacity credits. Variable RES power plants have low capacity
 credits. In order to meet reserve power constraints, the power model may require
additional thermal power (evidently from low capital-intensive technologies, such as
gas turbines), or may invest in pumped storage (depending on maximum possibilities
and costs), or may use import-exports more intensively. Similarly, flexibility services
depend on ramping possibilities of plants. Operation and investment in storage is
endogenous and compete against investment in flexible power plants. Costs of
developing reserves and flexibility service capacities are calculated and reflected onto
electricity prices, which are set at levels to recover total cost.
The model solution with endogenous flows over interconnections simulates common
balancing and/or market coupling and operation by load segment are synchronised
across the countries.

L.3.i. Environmental Policies

The PRIMES model computes emissions of air pollutants, such as SO2, NOx, PM, VOC,
from power generation and other types of industrial combustion.
End-of-pipe abatement is represented in the power model: auxiliary FGD, DENOx,
electrostatic filters, etc. are options for investment associated to existing or new
plants.
Usually in the scenarios, it is assumed that according to the Large Combustion Plant
directives, all new power plants shall be equipped with such end-of-pipe abatement
devices, and so they are included in new power plant investment.
Other provisions for old plants, such as transitory measures specifying maximum
operating hours for old plants, are represented in the model as constraints on specific
plants. Possibilities for retrofitting are endogenous.
The model allows for imposing ceilings on atmospheric emissions and associating a
shadow price (e.g. price of a SO2 permit). The provisions of the IPPC Directive on Best
Environmental Available Technologies are reflected upon the technical (hence economic)
policies characteristics of new technologies in all sectors. Similarly, other regulations are
handled, e.g. lighting.
CO2 emissions from industrial processes (non-energy related) are handled in relation
to industrial production of materials (e.g. cement). The model includes a marginal cost
abatement curve for reducing emissions and CCS investment for more drastic
emission cuts. Costs for CCS for processes are determined by accounting for capital
and variable costs. Electricity consumption for capturing is determined and adds to
total electricity demand. The PRIMES power/steam model is fully linked with the ETS
market simulation model of PRIMES and takes carbon prices from the latter. Sector-
specific carbon emission constraints as well as emission performance standards by
plant type can also be handled in the model.

Data Base L.4. Data sources of PRIMES power and steam model
The PRIMES power and steam model uses the following data sources:
• Inventory of existing power plants in all European countries (35 countries in
total) which is based on Platts and on ESAP; the data base includes power plant
name, location, name of owner, technology type, whether it is CHP or not, fuel
burning possibilities, gross and net power capacity, pollution control, date of
system commissioning, dates of possible refurbishment, indicative date of
decommissioning.
• Surveys of power plants under construction, which are qualified as plants with
exogenously fixed commissioning date; information by plant similar to inventory.
 • Eurostat statistics (complemented using data from ENTSOE, EURLECTRIC and
IEA) on power capacities, power generation by fuel or by plant type, fuel
consumption.
• CHP surveys by Eurostat (complemented by various sources and studies). The
CHP surveys are used in combination with plant inventory data and Eurostat
energy balances data to construct a complete statistical picture of cogeneration by
country. In this picture, industrial CHP including CHP on-site is fully represented;
this modifies Eurostat data because Eurostat does not show steam production and
fuel consumption by industrial on-site CHP separately (Eurostat shows only if CHP
is selling steam to steam markets).
• District heating surveys (from associations and industry). Surveys of industrial
boilers by sector (from industry)
• ENTSOe and national TSO/DSO data on electricity grids and interconnectors
• Data set of technical-economic characteristics of plant technologies that are
considered as candidates for investment. The data include: overnight investment
costs, fixed operation costs, variable costs, self-consumption rates, fuel efficiency
curves, ramping rates, technical minimum capacity, cost of start-up/shut-down,
emission factors (depending on fuel), CO2 capture possibilities if applicable,
technical maturity parameter (risk premium), distinction of scale (utility,
industrial, distributed). Technical data on possible configurations of electricity
and steam outputs of various CHP technologies. All technical-economic
characteristics of plant technologies change over time. Data are based on surveys
of several studies and databases, including IEA, VGB, etc.
• Potential of renewable energy sources for power generation. RES are classified by
technology and also in intensity classes by country. Potential data are used to
quantify cost-supply curves. Data come from various sources: ECN, DLR, GREEN-X,
Observ’er, etc. Wind velocity and solar irradiation data by country (TSOs and
private databases).
• Surveys on fuel prices, fuel procurement contracts, electricity prices by sector and
by component, grid tariffs, etc.
• Data on policies: feed-in tariffs, other RES supporting schemes, environmental
legislation etc.
• Underground CO2 storage possibilities by country.

L.5. Modelling of the EU power network and market

PRIMES has two options for computer running: a) full optimisation including all
European Internal
European countries simultaneously, b) separate optimisation per country with fixed
Market net imports, which are obtained after a full model running. The second option is used
in scenario variants in order to reduce computer-running time.
The regional model optimisation fully incorporates all existing and future
interconnection capacities (and Net Transfer Capacities). Electricity flows are
endogenous based on solving a DC linearized optimal power flow problem
simultaneously with all country optimisations. This means that an injection to a bus
(in the model a country node) is propagated to all interconnecting links. When
interconnection use is at capacity limit, marginal costs differ by node (country).
Interconnectors have limitations based on thermal capacity and feature reactance and
resistance.
 The optimal power flow (OPF) problem seeks to control generation/consumption to
optimise certain objectives such as minimizing the generation cost. Power flow
constraints can be approximated by some linear constraints in transmission
networks, and then OPF reduces to a convex program.
New interconnectors to be built in the future are projected based on planning by TSO
or on exogenous plans (e.g. Super-Grid, etc.). Power flows in DC lines are treated
differently than those in AC lines. The latter are linked with the countries’ net power
surplus or deficit by the power transfer distribution factors (PTDFs). On the other
hand, flows in DC lines are set by the solver when performing the optimization
without any link between each other (apart from being such that each country’s
balance is maintained). This reflects the fact that the link’s operator can control the
flow in a DC-link. Additional constraints mirroring Net Transfer Capacities or other
reliability constraints can be introduced. The choice of DC stems from anticipating a
possible evolution of the European grid, where a DC backbone could act as a highway
network that would bring renewable energy (produced in remote sites such as in
North Sea and in North Africa) to the big load centres of the continent.
This regional modelling approach can be also optionally used to study market
coupling policies or common balancing with coordinated management of
interconnectors. PRIMES can solve the power system model simultaneously by region
(alternative regional configurations are possible, user-defined), with endogenous
links between regions, which indicatively are the following:
1. Central Western Europe: France, Germany, Belgium, Netherlands, Luxembourg,
2. Central South Europe: Switzerland, Austria, Italy, Slovenia, Malta
3. Eastern Europe: Poland, Czech Republic, Slovakia, Hungary
4. Nordic and Baltic: Norway, Sweden, Finland, Denmark, Lithuania, Estonia, Latvia,
Kaliningrad
5. Iberian: Portugal, Spain
6. South East Europe: Romania, Bulgaria, Greece, Albania, Croatia, Bosnia &
Herzegovina, FYROM, Serbia-Kosovo-Montenegro, Turkey
7. British islands: UK, Ireland
External links (with exogenous or endogenous net imports): Russia, Ukraine,
Moldova, Belarus, Morocco, Middle East and North Africa (by country)

L.6. List of plant technologies

Solid fuel power technologies Gas firing power technologies
1. Steam Turbine Coal Industrial 1. Steam Turbine Gas Industrial
2. Steam Turbine Coal Conventional 2. Gas Turbine Gas Industrial
3. Steam Turbine Coal Supercritical 3. Gas Combined Cycle Industrial
4. Fluidized Bed Combustion Coal 4. Steam Turbine Gas Conventional
5. Integrated Gasification Combined Cycle Coal 5. Gas Turbine Combined Cycle Gas Conventional
6. Pulverized Coal Supercritical CCS post combustion 6. Peak Device Gas Conventional
7. Pulverized Coal Supercritical CCS oxyfuel 7. Gas Turbine Combined Cycle Gas Advanced
8. Integrated Gasification Coal CCS post combustion 8. Gas combined cycle CCS post combustion
9. Integrated Gasification Coal CCS pre combustion 9. Gas combined cycle CCS pre combustion
10. Integrated Gasification Coal CCS oxyfuel 10. Gas combined cycle CCS oxyfuel
11. Steam Turbine Coal Industrial 11. Internal Combustion Engine Gas
12. Steam Turbine Lignite Conventional 12. Peak Device Gas Advanced
13. Steam Turbine Lignite Supercritical 13. Small Device Gas
14. Fluidized Bed Combustion Lignite Biomass firing power technologies
15. Integrated Gasification Combined Cycle Lignite 1. Steam Turbine Biomass Industrial
 16. Pulverized Lignite Supercritical CCS post combustion 2. IG Biomass CC Industrial
17. Pulverized Lignite Supercritical CCS oxyfuel 3. Steam Turbine Biomass Solid Conventional
18. Integrated Gasification Lignite CCS post combustion 4. Peak Device Biogas Conventional
19. Integrated Gasification Lignite CCS pre combustion 5. High Temperature Solid Biomass Power Plant
20. Integrated Gasification Lignite CCS oxyfuel 6. Peak Device Biogas Advanced
Oil firing power technologies 7. Small Device Biomass Gas
1. Steam Turbine Refinery Fuels 8. MSW incinerator CHP
2. Gas Turbine Diesel Industrial 9. Internal Combustion Engine Biogas
3. Steam Turbine Fuel Oil Conventional Nuclear technologies
4. Peak Device Diesel Conventional 1. Nuclear fission second generation
5. Steam Turbine Fuel Oil Supercritical 2. Nuclear fission third generation
6. Fuel Oil Supercritical CCS post combustion 3. Nuclear fission fourth generation
7. Integrated Gasification Fuel Oil CCS pre combustion 4. Nuclear Fusion
8. Internal Combustion Engine Diesel RES technologies
9. Peak Device Diesel Advanced 1. Wind Power Low Resource
10. Small Device Light Oil 2. Wind Power Medium Resource
Boilers for industry and district heating 3. Wind Power High Resource
1. Fuel oil boiler 4. Wind Power Very High Resource
2. Coal boiler 5. Wind Offshore Power Low Resource
3. Lignite boiler 6. Wind Offshore Power Medium Resource
4. Diesel oil boiler 7. Wind Offshore Power High Resource
5. Gas boiler 8. Wind Offshore Power Very High Resource
6. Derived gas boiler 9. Wind small scale
7. Biomass boiler 10. Solar PV Low Resource
8. Waste energy boiler 11. Solar PV Medium Resource
9. Biogas boiler 12. Solar PV High Resource
10. Electric boiler 13. Solar Thermal
11. Hydrogen boiler 14. Solar PV Very High
Storage 15. Solar PV small scale - rooftop
1. Hydro with reservoir 16. Tidal and waves
2. Hydro Pumping 17. Lakes
3. Compressed Air Storage 18. Run of River
4. Hydrogen from electrolysis using RES 19. Geothermal High
5. Hydrogen from electrolysis to mix in gas supply Cogeneration technologies
6. Power to gas technology 1. Combined cycle with extraction
Industrial CHP Plants (by industrial sector) 2. Combined cycle with Heat Recovery
1. Coal 3. Backpressure steam turbine
2. Lignite 4. Condensing steam turbine with post firing
3. Natural Gas 5. Condensing steam turbine of large power plants
4. Derived gas 6. Gas Turbine with heat recovery
5. Oil 7. Internal combustion engine with cogeneration
6. Biomass 8. Others - backpressure steam for district heating
7. Waste 9. Fuel Cell
10. Very small scale Gas Turbine with Heat recovery
 Illustration of the power-steam production possibilities
of cogeneration technologies as represented in PRIMES

Possibilities of joined production of electricity and steam

by CHP plants (area ABDC)

L ine-1
Electricity
Output in MW
A Line-2

B Line-3

C
Line-4
D

Steam Output
in MW

L.7. Technology Progress

The technical-economic characteristics of technologies are assumed to change over
time (as a result of R&D and eventually economies of scale in mass production). The
rate of change of technical-economic characteristics over time is an assumption of the
modelling which may be altered depending on the scenario.
Unit Short-term Depending on the scenario, learning-by-doing and
Cost Cost supply curves economies of scale effects are introduced in the
quantification of technical-economic parameters for
both demand and supply-side technologies.
However, the PRIMES model does not include fully
Long run endogenous learning-by-doing mechanism in the
Cost curve power sector model (because of non-convexity
problems) but handles learning in the design of
Quantity of a renewable resource scenarios and by modifying parameters ex-post.
Technology progress is also assumed regarding the cost gap between technologies of
different scales, for example small-scale wind vs. large-scale wind parks. Scale
economics are also included through the representation of plants differently by
stylised scale (i.e. utility, industrial, distributed). A plant at industrial scale has worse
economics than plants at utility scale but benefits from possibilities of using CHP and
performing self-supply. Highly distributed plants, located at consumer premises, are
more expensive as they lack economies of scale, but bring benefits by avoiding grid
costs and losses. These mechanisms are simulated in the model.

L.8. Investment and Renewables

RES power Investment in new RES plants and operation of RES plants are fully endogenous. The
stochastic or variable RES (wind, solar PV, solar thermal, small hydro, tidal wave) are
represented as a deterministic equivalent power capacity: nominal capacity is
reduced according to the yearly resource availability rate and is assumed to operate
hourly at a share of nominal capacity according to exogenous resource availability
statistics by hour.
The hydro resources are considered dispatchable but constrained by yearly available
water flows: the model shows that they are used at peak hours until water constraint
is met.
The stochastic or variable RES get a “capacity credit” which is much lower than
nominal capacity. It varies by country and scenario depending on total deployment of
variable RES: capacity credit decreases with RES quantity deployed and differs by
country depending on assumptions about dispersion of RES sites. Capacity credits
enter the reliability or reserve power constraints: in case of large development of
variable RES, the model determines investment in low capital-intensive thermal
power plants (back-up) in order to meet reliability and reserve power constraints.
Thus costs increase and the competitiveness of variable RES decrease.
Flexibility of the power system to balance fluctuating RES is endogenously built and
reflected on investment as ramping possibilities differ by plant technology.
Large-scale storage is endogenous in the model (hydro-based pumped storage, air
compression and hydrogen-base storage): depending on economics, storage
simultaneously smooth load and accommodates transfer of RES energy from times
when RES availability exceeds load to times when RES is insufficiently available.
The model represents possibility for producing hydrogen from electrolysis and
blending hydrogen with natural gas (up to a maximum share of 30-40%). In case of
high RES development, hydrogen production, assumed to take place at off peak hours,
helps smoothing out the load curve, relaxing reserve power constraints and hence
allowing for more variable RES capacities
Regional power market operation under interconnection constraints also help
development of RES capacities and is simulated by the model (common balancing).
Cost of direct (shallow) connection of RES plants (wind, solar) with the grid are
assumed to be included in the unit investment cost of RES technologies; this is also the
case of offshore wind (assumed to develop at small distances from the coast). For
remote wind offshore, the model explicitly considers grid development for
connection. The model formulates impact of stochastic RES on power grid investment
and costs, distinguishing between offshore wind, onshore wind, solar PV and the level
of decentralisation.
Non-linear cost-potential curves for renewable resources (unit cost depend on
quantity and time) reflect difficulty of getting access to resource, availability of sites,
acceptance, grid connection difficulties, performance and for biomass land and waste
energy resource availability. Data on Potentials from: ECN (Admire-Rebus database),
DLR (database), Green-X, RES-2020, Observer, national sources, various studies and a
special data collection for biomass resources.
Investment in RES is projected on economic grounds as for any other technology. The
relative competitiveness of RES depend on technology progress (change of technical-
economic characteristics over time) and on policies supporting RES directly or
indirectly.
The unit cost of RES energy production is composed of the annuity payment for
capital (depending on WACC and risk premium), the fixed O&M cost and the variable
cost, where capital and fixed cost are divided by the number of yearly hours of full
capacity production, which depends on the resource availability rate. The capital part
of the cost of RES investment as perceived for decision making (not for actual
payment) is increased by a rate reflecting the non-linearly increasing cost-potential
curve, which may change by scenario according to assumptions about RES facilitation
policies.
Building a RES plant on an existing site (after RES plant decommissioning) is
considered to be cheaper than investing on a new site. Thus, the model captures the
fact that replacement of obsolete RES plants is much less expensive than building a
new RES plant.
For the biomass resources and commodities the model determines prices which span
the whole chain of activities and processes for producing and transforming feedstock,
reflect the possibly increasing cost of land use (for crops) and of collecting wastes and
price-setting components which reflect competition (for example pricing relative to

RES Feed-in tariffs: power purchasing agreements.

RES Obligation: in power generation a certain percentage
of electricity generated must come from RES (modelled as a
constraint).
RES blending obligation: fixed blending rate of biofuels for
transportation liquid fuels or for distributed gas.
Green certificates, Guarantees of Origin or generally
RES objectives: a RES shadow value is introduced (in
power generation and/or in other sectors) providing a price
signal to decision makers for getting benefits from RES; the
level of the RES value changes iteratively until the desired
volume of green certificate or of the RES target are
obtained; when iterating by keeping the same level of RES
value across sectors in a country or across countries, the
model can simulate RES certificate trading or other forms of
RES obligation exchanges between countries. Higher RES
values mirror more enabling policies for RES penetration.
RES facilitation policies: meant to include actions,
scenario-specific, which increase RES potential and make
cheaper the access to potential (reflected onto the
parameters of the cost-potential curves).
Note: RES as % of gross final energy is handled by PRIMES
as a target, either by sector, or country or EU-wide.
substitute fuels).
Direct RES subsidies reduce unit cost of capital or commodity prices (for biomass).
Feed-in tariffs are modelled as power purchasing agreements and they are subject to
budget constraint (by RES or overall). Investment in RES under feed-in tariffs is
modelled separately by simulating investment decision on individual projects based
on pay-back period calculated given the feed-in tariff amount. The model calculates
the probability of individual investment as a function of feed-in tariff level. This
mimics project finance accounting for RES investment. Payment for feed-in tariffs
under long-term PPA is fully accounted in the model. Revenues for these payments
are based on a special RES levy, applied directly on consumer tariffs. The amount of
the levy is endogenously calculated as a difference between RES feed-in tariffs and
marginal system costs (known from simulation of wholesale markets) since RES
displace fuels used by thermal generation.
Green certificates or renewable obligations are explicitly modelled as constraints,
which act on top of other support measures.
 When the entire PRIMES model is used to determine distribution of a renewables
target across sectors, including the power sector, the amount of generation from
renewables in the power sector is unknown. To find the distribution across sector, the
model introduces a shadow price of renewables obligations by sector, assumes that
this shadow price is equal in all sectors and varies its level until the overall RES target
is met. The shadow price of the RES obligation is called a RES-value (e.g. EUR/MWh
from RES) and can be interpreted as a virtual subsidy to generation from RES.
In power generation RES investment decisions are treated simultaneously with
system operation and reserve constraints, as well as with grid operations and costs,
which indirectly influence RES competitiveness.
RES curtailment is endogenous and depends on economics of the system without
affecting payments to RES owners under power purchasing agreements.
Illustration of using PRIMES to simulate conjoint development of renewables
and the EU network (super-grid) at a large scale

Non existing in 2005

DC Link for RES

FI

NO SE
Long
Distance EE
Offshore
LV
wind
DK
LT
KA CIS

IE
UK
NL
PL
BE
DE
Long LU CZ
Distance SK
Offshore AT
wind FR HU
SW
SI
HR RO
IT
BH YU

BG
FY
AL TK

PT ES
GR

AF
Solar Thermal, PV and wind in MENA

L.9. Investment and Nuclear Energy

Nuclear Investment in nuclear power is treated as an economic decision. Nuclear deployment
depends on electricity demand, load profiles, economic features of competing
technologies and carbon prices. Nuclear decisions, taken together with all other
power plant decisions, fit within least cost capacity expansion to a long-term horizon
 (under perfect foresight) and within least cost unit commitment and is influenced by
policy drivers for example by carbon prices.
Investment decision on nuclear distinguish between:
• Extension of lifetime of an existing plant (involves investment cost lower than
for a new plant)
• Building a new nuclear plant on an existing site, if such possibility exist
(investment cost is lower than for a new site)
• Building a new nuclear plant on a new site (Greenfield development)
The unit cost of nuclear plant investment differs by country depending on economies
of scale experienced in nuclear industry: it ranges from first-of-the-kind investment
costs for countries that may invest in nuclear for the first time to investment cost
levels corresponding to high economies of scale. The unit cost of investment depends
on the nuclear technology: second, third and fourth generation technologies are
represented in the model database. Regarding new nuclear plants, In PRIMES, nuclear
second refers to generation II reactors until 2015 and will include after 2015 the
commercially most advanced reactors (e.g. EPR, AP, VVER 1200) currently
summarized under generation III and III+; nuclear third refers to the remaining set of
generation III+ reactors and nuclear fourth refers to generation IV reactors.
Technology progress over time is represented differently by nuclear technology. The
unit cost of investment take into account costs for future decommissioning (15%
provision). Variable and fuel costs of nuclear power take into account waste recycling
and disposal costs.
The lifetime of old nuclear plants is set as specified in their license and is extendable
upon investment. New nuclear plants are supposed to have lifetime of 40 years,
extendable after investment.
The model can represent the following possible policy constraints on nuclear
investment:
• No nuclear in the future
• Phase-out of nuclear
• Fixed decommissioning dates for specific plants
• Permission of extension of lifetime (as economic decision or as a
decided policy)
• Permission of investment only on existing sites
• Upper bounds on nuclear expansion
• No constraints on nuclear expansion
To reflect growing cost of developing new nuclear sites, the model includes site
specific cost elements, which differ by country and evolve over time (or set as a
scenario specific assumption). These costs are based on cost-potential curves, which
apply only for Greenfield investment and are nonlinear with increasing slopes.
Policies aiming at higher nuclear are represented by shifting the nuclear cost-
potential curve to the right (lower cost for equal potential).
The PRIMES model database includes a detailed inventory of nuclear power plants
that are in operation in Europe and their characteristics and includes new nuclear
projects, which are under construction or in consideration.

L.10. Investment and CCS

CCS CO2 capture, transportation and underground storage (CCS) is one of the possible
means of reducing CO2 emissions from combustion of fossil fuels in power plants and
in industry. Driven by emission reduction targets or by carbon pricing, CCS (if
considered available at a certain period and countries as part of scenario
 assumptions) competes with other means, such as carbon free power generation
(renewable energies, nuclear), the fuel switching towards low emitting forms and the
reduction of energy consumption.
The power plants with carbon capture are more expensive in terms of capital
investment and operation costs than similar plants without carbon capture. Moreover,
their net thermal efficiency is much lower, since carbon capture needs electricity to
operate.
The costs of transporting and storing CO2 are modelled through non-linear cost curves
by country, bounded by storage potential (set exogenously for each scenario per
period). Costs increases with quantity stored. Data come from TNO and JRC.
It is assumed that CO2 transportation and storage are offered by regulated monopolies
operating by country (CO2 exchanges between countries are mot modelled in the
current model version). Investment is considered as exogenous and varies by
scenario. Transportation and storage activities operate under strong economies of
scale, bear very high fixed costs (and small variable costs) and face high uncertainty
about future use of infrastructure. Prices for transportation and storage services are
determined based on levelized total development costs and investments over time on
an anticipated cumulative demand for the service.
Public acceptance issues and other uncertainties are expressed through parameters
shifting the cost-supply curve to the left and up (making more expensive the service
and lowering potential).
Scenarios involving delays in CCS development may be simulated by introducing
particularly high storage and transport costs for a limited period.
The pilot CCS plants envisaged for 2020 are assumed to have reserved specific sited
for CO2 storage at rather short distances with small marginal costs for storage.
The CCS investment decisions are integrated within the PRIMES sub-model on power
and steam generation. The CCS technology for power plants is represented in two
ways: a) as typical new power plants enabled with CCS considered as candidate for
investment, b) as auxiliary technologies candidate for retrofitting existing power
plants or plants built (endogenously by the model) without initially having the CCS.
This flexible representation allows assessment of various policy options, as for
example the “capture-ready” options or mandatory CCS measures.

L.11. Financial and Pricing Model for Electricity, heat and steam
The financial power/steam model operates after the run of the power/steam
Financial and
optimization model.
Pricing Model
The first step is to calculate in detail all costs of power and steam/heat production,
based on the results of the power/steam optimization model. The costs, separately for
electricity, steam and heat, are calculated as time series and include:
a) Capital investment costs: total investment expenditures and annuity payments
for capital calculated using a weighted average cost of capital, which mirrors
business practices, and differ by plant scale, i.e. utility, industrial and
distributed. Capital costs of non-yet amortized old plants are included.
 b) Variable costs which include variable operating, fixed maintenance, fuel costs
and payments for fuel taxes, emission taxes and ETS auction payments12.
c) Costs of electricity supply and trading (by country and by voltage category)
d) Total system payments for direct subsidies or power purchasing agreements
(feed-in tariffs) for renewable plants.
e) Other costs including for public service obligations if applicable.
f) Grid costs, separately by grid type, calculated according to a regulated asset
basis methodology, which includes capital costs of old infrastructure, cost of
new investment and operating/maintenance costs.
The electricity prices in PRIMES are calculated in order to recuperate all costs,
including capital and operating costs (and possible stranded investment costs), costs
related to schemes supporting renewables, grid costs, and supply costs.
The next step aims at determining the electricity (and steam/heat) prices by category
of customer (sectors and sub-sectors of demand). Each customer type has a load
profile, which is calculated in the PRIMES demand models, and partly the customer
may be self-supplied. The aim is to allocate variable, fixed, and capital costs as well as
grid and other costs to each category of customers as it would be resulting from a
well-functioning market in which suppliers would conclude efficient and stable
bilateral contracts with each customer category based on the specific load profile of
the customer.
To do this, the following calculation steps are performed:
a) Simulation of virtual wholesale energy-only markets by country in order to
estimate marginal system prices reflecting fuel marginal costs.
b) Calculation of marginal long-run costs of the system, including capital costs by
vertical time zone of the marginal plant, in order to allocate capital costs to
customer-types according to their load profiles; this step can be used to
estimate capacity remuneration which is eventually needed to address the
“missing money problem” of the energy-only wholesale market approach
performed in the first step.
c) Allocation of costs of energy losses in transmission and distribution to
customer types according to their connection to voltage categories depending
also on the share of demand covered by grid electricity (self-supply is
excluded).
d) Matching of load profiles of customer-types with the duration curve of system
marginal prices including long-run capital cost components with customers
sorted in descending order of their load factor mimicking bilateral contracting.
Calculation of payments by customer category and comparison to total
recoverable costs.
e) Calculation of fixed capital/maintenance costs not recovered from prices
determined in previous step. Inclusion of possible other costs, such as costs
for public service obligations, ancillary services, etc. Allocation of these
remaining fixed costs to customer categories using a Ramsey-Boiteux method:

12In case of ETS with free allocation of allowances, the cost calculation considers that allowances have an
opportunity cost, based on ETS carbon prices. Depending on the scenario, the model allows for assuming
that part pf this opportunity cost can be passed through to consumer tariffs, depending on the intensity of
market competition.
 a mixed complementarity problem, which allocates fixed costs at inverse
proportions of assumed price-elasticities by sector. At this stage, cost mark-up
factors (positive or negative) are exogenously added to reflect different
market circumstances and regulatory practices, including price regulations,
pricing below total recoverable costs (which is the case in some countries),
pricing above recoverable costs due to market power, practices of cross-
subsidization between consumer categories, etc.
f) Calculation of revenues of RES supporting schemes and comparison to
payments, in order to determine a RES levy to be paid by customers. Revenues
of RES are estimated based on short-term marginal system costs calculated in
the first step of the process, to reflect marginal fuel costs displaced by the use
of RES. The level of the RES levy may vary by customer category depending on
policy assumptions.
g) Grid costs are recovered from grid use tariffs determined by customer
category, depending on connection to voltage category. The grid tariffs are
calculated as levelized long-term prices as required to recover regulated asset
basis, using an assumed discount rate which mirrors regulation practices in
the different countries.
h) Calculation of final end-user prices by sector based on prices of virtual
bilateral contracting, allocation of remaining fixed costs, grid and losses tariffs,
RES levy, etc. Recovery of total system budget is ensured depending on
assumed mark-up factors on costs.
i) End-user excise tax and VAT rates are added to electricity prices. These prices
are transmitted to PRIMES demand models.
Prices of distributed steam and heat follow a similar, but simpler, methodology.
Recovery of costs of steam and heat distribution networks is based on tariffs
calculated as levelized prices using regulated asset basis. The price of steam produced
by CHP is calculated based on opportunity costs: they reflect the marginal cost of
avoided boiler use.

L.13. Simulation of Oligopoly Competition in Electricity Markets

Market To study regulatory issues related to market competition at national and EU levels,
the PRIMES model uses a modified version of the power/steam model.
Competition
Model Version As mentioned above, the standard power model solves least cost optimization, which
corresponds to perfect long-term market competition. Electricity prices computed on
this basis reflect long-run marginal costs. In these conditions, investment schedules as
projected by the model exactly match demand and reserve and flexibility
requirements and investors fully recover capital costs. Studying regulatory policies
requires dealing with possible market imperfections and situations where investors
may not fully recover capital costs unless exercising market power or receiving
capacity remuneration.
The market competition version of the power/steam model introduces competition
between generation and supply companies, which are explicitly represented as
owners of plants and potential load serving entities in some countries. The number of
companies actually competing in each national market (optionally in each regional
market) indicate the intensity of competition. Each company is supposed to act
according to Cournot behaviour under conjectural variations. The model finds a Nash
equilibrium when solved. Conjectural variations include parameters, which can
represent a variety of competition regimes, ranging between quasi perfect
competition, supply function equilibrium, pure Cournot and monopoly competition.
Investors consider plant investment as a project financing problem and decide upon
implementing the investment if the internal rate of return exceeds a certain threshold.
The return depends on earnings based on system marginal prices and on eventual
capacity remunerations. Uncertainty factors mirroring poor predictability are
introduced in the investment appraisal calculations, which can be varied by scenario.
The model does not allow unmatched demand and supply, as it assumes that system-
driven open-cycle turbine plants become available in case of market failure to cover
demand; such investments are paid as part of system charges.
Trade of electricity between countries is driven by differentiation of system marginal
prices between countries (or regions) and the model formulates arbitragers (traders)
which take profit from price differentiation by transferring electricity between
country nodes as much as allowed by the interconnecting system, which is
constrained by the Kirchhoff law, and possible regulatory/reliability restrictions.
Transmission system operators are also modelled as owners of interconnectors with
behaviours seeking rents from interconnection congestion. Traders and suppliers are
formulated to anticipate such congestions by seeing the marginal rents.
Electricity prices are not exactly reflecting long-term costs as they depend on market
competition intensity; in other words cost mark-ups are endogenous in this model.
Hence, demand response is also endogenous and this constitutes in fact the main force
of mitigating market power. Demand is modelled as having varying price flexibility by
sector.
The market competition power model of PRIMES can handle asymmetric capacity
adequacy policies in the EU as well as harmonised policies. The model has been
successfully used in studies on capacity remuneration schemes for the European
Commission.
 L.14. Power sector sub-model of PRIMES Version 6
The version 6 of PRIMES includes a significantly enhanced version of the power sector
Enhancements in sub-model of PRIMES. The improvements are mainly twofold:
version 6 of the
a) representation in higher detail and resolution of the existing fleet of power
power sector plants in Europe and so capturing in a better way the projection of
model of PRIMES decommissioning, refurbishment and new constructions;
b) improvement of the model capability in simulating unit commitment in the
presence of high contribution by variable renewables and so capturing in a
better way the system requirements for operation of fast-ramping power
resources (flexibility) and the possible sharing of such resources within the
EU internal market based on cross-border trade and market coupling.
The new developments make the model considerably better placed to study policy
issues for the internal market, the integration of renewables and the simulation of
investment behaviour. Recent experience from the market suggests that investment in
power plants relies less than before on theoretical long-term optimality of costs.
System-depending operational restrictions deriving from penetration of variable RES
imply forced operational cycling of plants. Ignoring them in economic appraisal of
investment would be a serious drawback. In addition, the refurbishment options are
highly influenced by increased regulation regarding the air pollution emissions, for
fossil fuel plants, and by security regulation, for nuclear plants.
The enhancement of the model addresses these main issues in a considerably
improved manner, due to higher resolution of load variation (daily and seasonal) and
to higher resolution of the handling of old plants. More specifically:

a) The model data decomposes load variation into 120 different time segments
per annum. They cover typical days by season, for working days and for
weekends and holidays. The decomposition took into account the variability of
wind and solar by region (country level at least). The design of the time
segments, based on statistical algorithm, aimed at capturing the variance of
load and simultaneously the variance of solar and wind. Therefore, the
distribution distinguishes typical days with for example high solar irradiation
from similar days with low irradiation, and similarly for wind intensity. The
data represent variability of solar and wind on an hourly basis and the
aggregation by time segment takes care to capture the simultaneity of
variance with load. This work drew on a vast data collection of hourly load
curves and hourly wind and solar resources on geographical scales.

b) The enhanced model performs unit commitment (dispatching) in considerably

more sophisticated way than in the past, as the time resolution allows
including technical details on cycling operational constraints of power plants,
such as ramping rates, minimum power levels for stable generation, start-up
and shutdown costs, etc. These are important to capture risk of over-
generation (with possible curtailment of renewables and low levels of
wholesale prices), reliability of ramping (flexibility resources) and backup
(reserve) power. Constraints also reflect reserves for ancillary services, which
are also important for real time balancing in the presence of high variable
renewables.

c) As the model operates unit commitment (dispatching) over the entire

European network, it captures the possibilities of sharing power resources
among the system control areas, which depends on availability of
interconnectors and the possible constraints arising from net transfer capacity
restrictions, as opposed to flow-based allocation of interconnection capacities.
 Thus, the model simulates different regimes of market coupling and their
impacts on dispatching, costs, and efficiency.

d) Obviously, investment in new power plants, as well as in storage systems,

depends in a much more accurate way on the sophistication of dispatching
simulation, within the enhanced model.

e) In addition, the enhanced model includes in more detail the carious storage
technologies, including pumped hydro storage (with distinction between
mixed pumping and pure pumping, which is a new feature), batteries, air
compression and a number of power-to-X technologies, where X stands for
hydrogen, gas, synthetic liquids, etc. These latter technologies act as storage
systems indirectly, and obviously also as source of electricity-based (thus
possible renewables-based) new energy carriers.

f) The model applies a sophisticated Benders decomposition algorithm to

perform sequence of unit commitment over the entire European network
(including simulation of wholesale markets, coupled or not) and investment
decision cycles. The new algorithm is computationally more efficient and
allows introducing uncertainty factors influencing specifically the investment
behaviour. Limited foresight also applies, optionally, in the investment
behaviour.

g) As the new model version uses Benders decomposition, the investment

decision runs separately from unit commitment. Model options are used to
specify the time horizon of foresight influencing investment decisions with
additional parameters handling uncertainty and anticipation, which allows for
the representation of non-perfect foresight if required. The previous version
of the model was solving simultaneously optimal expansion and unit
commitment and thus it was impossible to introduce sophistication in the
investment behaviour to capture influencing factors more realistically.
Similarly, the decomposition allows capturing in more detail the specific
investment behaviour subject to feed-in tariffs or other similar mechanisms
including auction-based contracts for differences or power purchase
agreements. This is also among the enhancements of the new model version.

Although the database of the power model includes all existing power plants in
Europe on an individual basis, the simulation aggregates them in categories.
Previously the categories reflected the main power technology and fuel. In the
enhanced version of the model, the aggregation considered several dimensions,
including location, size, technology, age of the plant, and fuel.
Thus, the number of different existing plants in the model increased considerably and
the capturing of specific features improved a lot. For example the model now includes
in all countries the large power plants one by one (individually) and not aggregated as
before. The aggregation mostly concern small plants. The new classification of plants
also puts more emphasis on disaggregating the categories by each industrial sector
and in addition by technology, age, and fuel. The aim was to capture industrial
cogeneration in an improved manner from a sectorial perspective. This was necessary
for the enhanced industrial model on boilers and cogeneration.
The enhanced model further allows for the inclusion of cyclical operation constraints
of plants, which differ by technology. As we perform unit commitment over
chorological sequence of load – at hourly resolution- the model is able to include
ramping constraints, minimum up time, minimum stable generation levels, etc. The
model further now fully includes flexibility and secondary reserve requirements at
system level; hydro power, imports and gas are the main providers of flexibility.
 Operational costs of cycling operation (default) compared to forced provision of faster
ramping and cycling imply higher operation and maintenance costs, which are
included in the model. This kind of operation mainly applies to GT, CCGT and other
plants operating under AGC control of the TSO.
The benefits of investing in flexible plants arise for system reliability purposes, which
are represented as reserve requirement; the monetary benefits for investors depend
on specific remuneration of the service. Within fully optimal model resolution, such
remuneration is implicit as full cost recovery is assured in the modelling.
The model can be adapted in order to be able to represent failures, which can incur
with such systems. It is important to note that flexibility, as well as reliability, have
public goods features, this means that private investment can be subject to free riding
by competitors not investing.

L.15. Special version of power sector model for the Internal

European Market: The PRIMES/IEM model
Special version of L.15.a. Introduction to PRIMES/IEM model
The modelling analysis with the PRIMES/IEM aims to simulate in detail the sequence
the power sector
of operation of the European electricity markets, namely the Day-ahead market, the
model utilized as a Intra-day and balancing markets and finally the Reserve and Ancillary Services
market operation market or procurement. The PRIMES/IEM modelling suite consists of four main
simulator with models:
high resolution  A Day-Ahead Market simulator (DAM_Simul), which simulates the operation of
the day-ahead market and is based on the EUPHEMIA algorithm13.
 A Unit Commitment simulator (UC_Simul), which simulates the scheduling of
units occurring real-time, considering fully technical limitations of power
plants.
 An Intra-Day and Balancing market simulator (IDB_Simul), which simulates the
operation of the market for balancing services and the settlement of deviations
which occur between the real time scheduling of units (output of UC_Simul)
from the day-ahead (output of DAM_Simul).
 A Reserve and Ancillary Services market simulator (RAS_Simul), which
simulates the reserves and ancillary services market procurement.
The PRIMES/IEM covers all EU 28 Member States individually, in detail. It also
represents Norway, Switzerland and the Western Balkan countries, in an aggregate
manner, in order to account for exchanges of energy between EU and these countries.
PRIMES/IEM disaggregates the interconnection network, and considers more than
one node for each country, in order to represent in-country grid congestions. The
assumptions about the grid within each country and across the countries change over
time, reflecting an exogenously assumed grid investment plan. Existing power
capacity of lines and new constructions are based on ENTSOE data and the TYNDP.
Technical characteristics of transmission lines (thermal limits and admittance factors)
have been collected from TSOs.
The power market simulators of the PRIMES/IEM can be calibrated to projections of
the standard PRIMES model for any specific scenario, and can run for any year of the
projection (usually 2015 to 2050 by 5-year periods). Inputs from a scenario include:

 Load demand (hourly), power plant capacities, net imports with countries
outside of EU28, capacity of the transmission lines and net transfer values NTC
values.

13EUPHEMIA (Pan-European Hybrid Electricity Market Integration Algorithm) is the single price coupling
algorithm used by the coupled European PXs
  Fuel prices, ETS carbon prices, taxes, etc.
 RES generation, however the simulators of PRIMES/IEM can determine
endogenously curtailment.
 Potential of hydro production (for hydro reservoirs). Constraints on water
availability have been applied on a daily basis in PRIMES/IEM. We have
separated mandatory hydro-lakes production (due to excess water and other
uses of water) from hydro-lakes production at peak load times. This is an
important distinction for the bidding behaviour of lakes.
 Heat or steam serving obligations of the CHP units whose main product is heat
or steam rather than electricity (industrial CHP and small CHP units exclusively
used for steam and heat).
 Other restrictions derived from specific policies, e.g. operation restrictions on
old plants, renewable production obligations, and, if applicable, support
schemes of renewables, biomass and CHP.
The PRIMES/IEM incorporates a detailed database by plant, with disaggregated
technical and economic data for each plant in order to be able to represent cyclical
operation of plants, possible shut-downs and start-ups. Its database also includes
detailed data on the technical possibilities of plants to provide ancillary services. The
ancillary services represented in PRIMES/IEM include Frequency Containment
Reserve (primary reserve), Automatic Frequency Restoration Reserve (secondary
reserve – Automation Generation Control or AGC), Manual Frequency Restoration
Reserve (spinning tertiary reserve) and Replacement Reserve (non-spinning tertiary
reserve). Relevant data have been collected from the national TSOs.
Finally, the PRIMES/IEM represents typical 24-hour days, which are distinguished by
season, and by working days or holidays (and weekends). For example, a typical day
could be a working day in winter.

L.15.b. Modelling procedure

First, the Day-Ahead Market simulator runs (DAM_Simul), and yields with a unit-
commitment schedule of power plants, including demand response and schedule of
flows over interconnectors.
After the simulation of the Day-Ahead market, we use a “Random Events Generator”
tool (developed as part of the PRIMES/IEM specifically for the purposes of this
analysis) to generate a set of random events (experiments). The purpose of this step is
to artificially introduce a deviation between the day-ahead forecasts (on load, RES
generation, availability of plants etc.) and what is occurring real-time.
Considering these deviations, we run a unit commitment simulation with the
UC_Simul, which is similar to the day-ahead simulation, with the difference that it
includes constraints on the technical operation capabilities of plants. The outcome
compares to the day-ahead simulation outcome, and the difference serves as a best
forecast of deviations by the market participants.
The next step is the financial settlement of the deviations between the day-ahead
schedule and the UC_Simul schedule, and is undertaken with the Intra-Day and
Balancing Market simulator (IDB_Simul). The IDB_Simul runs and bids for deviations
are generated.
The next step is the simulation of the market or procurement for reserve and ancillary
services, conducted with the RAS_Simul tool. The simulation takes into account the
commitments in the previous stages, and determines the offerings and remuneration
for reserve and ancillary services, given exogenously set reserve requirements.
 The table below gathers the steps of the modelling work performed. The following
paragraphs describe in more detail each modelling tool of the PRIMES/IEM and the
methodology followed in the simulations.
Steps of modelling work performed with the PRIMES/IEM modelling suite

Steps of the
PRIMES/IEM Process Output
simulations
Plant and interconnectors
Step 1: Running of the Simulation of the DAM simultaneously for all EU
operation schedule (DAM
Day-Ahead Market countries. Basis is a PRIMES scenario (capacity,
schedule), and its
simulator demand, must-take generation, etc.).
financial settlement

Step 2: Generation of Generation of experiments (events), with deviations

52 cases (random events),
experiments with the from the PRIMES scenario for wind and solar
and respective
Random Events generation, demand, availability of plants and
Generator frequencies
interconnections
Revised plant and
Step 3: Running of the For each random event, UC simulation considering interconnectors operation
Unit Commitment all technical constraints of plants schedule (UC), deviations
simulator from the DAM schedule

Step 4: Running of the Set-up of market to settle deviations from DAM Financial settlement of
Intra-day and defined with the UC simulator. Eligibility by plant to deviations and revised
Balancing Market bid in the IDB is determined hourly, based on schedule for operation of
simulator output of UC. units and interconnectors
Step 5: Running of the Settlement of exogenously set reserve
Reserves and Ancillary The remuneration of the
requirements, considering residual capacity after
Services market or resources for providing
procurement IDB settlement.
reserves
simulator
Calculation of financial balances (revenues and costs) for each generator, load
payments (payments by consumers) and payments by the TSOs. Calculation of
Step 6: Final cost-
unit cost indicators (e.g. for reserves, etc.). Calculation of expected values of the
accounting
outcomes, as average of results by case (random event), weighted by the
frequency of each case.

L.15.c. Day-ahead market simulator (DAM_Simul)

The DAM_Simul algorithm consists of a set of equations which replicate the
EUPHEMIA algorithm. The core parts of the algorithm of the DAM_Simul is a balancing
equation, which regulates the inflows and outflows in each node, and the objective
function which is such so as to maximize the social surplus. These are complemented
by network equations, which allow to simulate a flow-based allocation of
interconnection capacities. The model also includes equations related to operational
limitations, which guarantee that all plants should offer energy below their maximum
capacity or that over-the-counter arrangements (nominations of energy) should be
respected. DAM_Simul runs for all EU countries simultaneously, with every country
representing a node, and determines market clearing by node and interconnection
flows. DAM_Simul produces unit schedule, use of interconnectors and SMPs. It also
performs the financial settlement of the DAM. DAM_Simul draws techno-economic
and other data from the standard PRIMES model database, including:

 heat rate per plant,

 fuel prices,
  CO2 emission coefficient per plant,
 cost parameters per plant (fixed and variable),
 power network topology and technical characteristics,
L.15.d. Network representation in DAM
In the DAM_Simul, every country is represented by a single bus and the network
includes all current AC and DC interconnections, as well as known investments
according to the TYNDP. The model simulates optimal flow-based allocation of
capacities across interconnections. The flows are restricted by the first and second
Kirchhoff laws and by administratively defined Net Transfer Capacity (NTC)
limitations, applying to pairs of adjacent countries.
Depending on user-defined options, flows may be further restricted by over-the-
counter arrangements, such as the Available Transfer Capacity (ATC) restrictions
implying limitations not only by the NTCs but also by the amount of capacity engaged
for cross-border nominations.

L.15.e. Bidding of power plants

Energy offers by plant in the DAM_Simul can be of various types, including hourly
orders, flexible hourly orders, block orders and complex orders (defined as in
EUPHEMIA). We assume that plants bid in the markets (if bidding is allowed in the
option under analysis) at the level of their marginal cost (which is taken from the
PRIMES database) plus a scarcity mark-up. The mark-up is derived according to a
scarcity bidding function (defined by plant), which takes into account hourly demand,
plant technology and plant fixed costs. The use of the scarcity bidding function serves
as a means of mimicking the strategic bidding behaviour of plant owners in an
oligopoly. Regarding hydro-reservoir power plants (lakes), we assume that part of the
generation of lakes (mandatory generation) makes zero biddings, simulating the
energy that needs to be used in order to avoid overflow and/or for irrigation
purposes (particularly relevant for southern EU countries). Only the non-mandatory
part of lakes is bidding in the day-ahead market. Bidding is determined through a
scarcity bidding function (similar to all other power plants), which takes as basis the
marginal cost of gas generation and adds a mark-up that reflects scarcity during peak
hours and water availability.
It should be noted that no negative bidding is assumed in the simulations, for any type
of power plants. This has effects for the results of the simulations, mainly in the
analysis of results regarding priority dispatch of variable RES. In particular, in policy
options that assume no priority dispatch of variable RES, these capacities make offers
at very low prices, as their marginal cost is close to zero, and are therefore very
competitive. In case negative bidding was allowed, and given the increased variability
introduced in the dispatching schedule due to the high shares of variable renewables,
some inflexible units would bid negative prices in order to e.g. maintain a minimum
operation level. In these hours, the close-to-zero bids of RES would seize to be
competitive and thus RES generation would be curtailed. Without negative bidding,
RES are still curtailed (if no priority dispatch is assumed), but not as often as they
would if negative bidding was allowed.
Depending on user-defined options, the model can handle different assumptions
regarding the existence of bidding zones. They can vary from national, to regional or
to a fully integrated EU markets with flow-based allocation of the entire capacity of
interconnections, as if there was a single bidding zone.

L.15.f. Modelling of nominations

Depending on user-defined options, the simulation with the DAM model assumes
nominations of energy, i.e. scheduled and fixed generation by power plants. In
particular, in the options that this assumption is activated, for example the generation
of nuclear and solids-fired power plants is treated as nominated. Part of this
nominated energy can be allocated to contribute to the fulfilment of cross-border
trade contracts (cross-border nominations). In-country nominated energy does not
participate in the DAM solution. Cross-border nominated energy is subtracted both
from supply of the country of origin and demand of the country of destination, while
ATC value is reduced in their borders accordingly.

L.15.g. Modelling of priority dispatch

Depending on user-defined options, the DAM simulation can handle priority dispatch
for certain power generation technologies, such as RES, industrial CHP units, district
heating plants and others. In order to simulate priority dispatch for certain capacities,
we assume zero biddings for fixed hourly amounts of generation of these capacities
and introduce a very high penalty for curtailment, high enough so as no curtailment is
occurring when we assume priority dispatch.

L.15.h. Modelling of demand response

Demand response is modelled so as to allow for endogenous shifting of demand
quantities among hours. The shift comes as a response to price signals, thus shifting
demand from hours of peak prices to hours of low prices, leading to a smoother
demand curve. The modelling of demand response uses stepwise functions that
associate amount of demand response to relevant costs, exhibiting decreasing returns
of scale (i.e. the higher the amount of demand response the higher the cost).
Consumers can bid for demand response, with bidding quantities being subject to
potential and bidding price reflecting the stepwise costs.

L.15.i. Day-ahead market simulator version with Unit Commitment

(DAUC)
The DAM model includes as an option the possibility for optimising simultaneously
offering of energy and reserves. When this option is activated, the DAM model
algorithm includes additional equations:
a. Equations that represent plant-related technical constraints (e.g. technical
minimum and maximum, ramping capabilities, minimum up hours of operation,
etc.). This implies that the DAM simulator in these options includes the same
set of equations as the Unit Commitment simulator. The fact that the DAM
model becomes similar to the UC model, implies that deviations between the
day-ahead market scheduling and the real-time scheduling simulated by the UC
are the least possible, and attributed solely to random events occurring
unexpectedly real-time.
b. The equations of the Reserve and Ancillary Services Procurement model. In
practice, when the co-optimisation option is activated, it is as if the two models
run simultaneously.
The problem of co-optimising energy and reserves is a mixed integer problem, with
binary variables reflecting plant operation status. In order to derive the SMP, we
perform a second run, after fixing variables to the integer solution, and relaxing the
integer constraints, allowing them to be linear. This is necessary, as the SMP should be
able to take non-integer values.

L.15.j. Random Events Generator (optional)

Using a random events generator is basically the modelling way to represent the
difference between “what is projected to happen” the day-ahead and “what is
happening” real-time. The simulation of the day-ahead market starts by taking as
given the hourly demand, hourly renewables generation and hourly availability, as in
a PRIMES scenario. This can be seen as the projections of the TSOs for the following
day, based on which the day-ahead market clears. The random event generator,
generates a set of deviations for these projections trying to mimic the deviations that
could occur in reality between what is expected to happen the day after and what is
indeed happening, as for example for:
c. Deviations in the load pattern (demand)
d. Deviations in the generation of variable RES, namely wind and solar
e. Unexpected unavailability of large power plants (equivalent to having
unexpected outages)
f. Unexpected reduction in the net transfer capacities (NTC) (equivalent to having
unexpected loss of transmission lines)
With the above as variables, the random events generator builds a large number
stylised cases, or experiments, each of which is assigned with a probability of
occurrence. The assumed level of deviation of the variables and the probability of each
experiment are based on expert judgement. Days of the year are classified to clusters
according to their characteristics (season, whether it is working day or holiday), and
each experiment regards a specific day cluster.
The simulations with the Unit Commitment simulator and the modelling of
settlements for deviations between the Day-ahead market and the Intra-day market
(with the Intra-Day and Balancing market simulator) are conducted for all
experiments, and final results are reported after weighting the results for each
experiment with the assigned probabilities.

L.15.k. Unit Commitment simulator (UC_Simul)

The Unit Commitment simulator mimics real-time operation, where all uncertainties
regarding technical constraints as well as forecast deviations have been resolved.
Note that we assume that all uncertainty is resolved in one step, i.e. we do not perform
runs of the UC_Simul for different points in time between the DAM closure and the
real-time (as if continuous IDM were operating).
In the UC_Simul, generators compete simultaneously for providing energy and
reserves14. The model takes as inputs exogenously defined reserve requirements, the
outcomes of the Random Event Generator on deviations to forecasted (in the day-
ahead simulation) demand, renewables generation, availability of power plants and
NTC values, and the dispatch schedule and bidding according to the DAM simulator.
The amount of capacity that is bound for reserve purposes according to the
optimization with the UC model is not participating at all at the clearing of the intra-
day and balancing markets that follows. Therefore, in the modelling reserve refers to
the capacities that participate in the reserve markets or procurement after the
clearing of the intra-day and balancing markets.
The model runs for the pan-European electricity network, following the same
modelling of power flows as in DAM_Simul. At this stage, network representation is
more detailed in the UC_Simul than the DAM_Simul, in that for some countries it
includes more than one node. In this way, in-country network limitations, which are
of high importance not only for the flow of power within the country but also for the
international flows, are taken into consideration.
The UC model uses the DAM solution as initial condition. The solution of the UC_Simul
regarding dispatching of units differs from the solution of the DAM model due to:
a) The consideration of technical constraints of power plants in the UC_Simul.
In particular, in the UC_Simul all technical constraints that generators
encounter in real-time operation are represented from respective

14 At this point it is important to clarify what is considered as reserve in the modelling. Demand for reserves

is predetermined and exogenous.

 equations, namely the maximum hours of operation above technical
minimum, ramping constraints, minimum up and down time constraints.
Plant specific start-up and shut-down costs are also included.
b) The consideration of demand for reserves, in other words, the simultaneous
optimisation of offers for energy and reserves. UC_Simul includes equations
that represent the technical limitations of each power plant for providing
ancillary services.
c) The deviations on load, renewables generation, availability of plants and
NTC values introduced by the Random Events Simulator.
Bidding by generators for energy follows the same logic as in the DAM_Simul, i.e.
generators bid strategically according to a scarcity bidding function. Bidding for
reserves varies depending on user-defined options. The user may define
administratively regulated offers for reserves, both in terms of quantities and in terms
of payments for ancillary services (based on procurement and contracts), or
opportunity cost bidding for reserves by generators, i.e. on the value that generators
lose by binding their capacity for reserves instead of bidding this amount of capacity
in the energy markets. The opportunity cost is calculated taking into account the
hourly SMP price of the DAM solution. In particular, if we denote 𝑂𝐶𝑛 the opportunity
cost of plant 𝑛, 𝐺𝑒𝑛𝐷𝐴𝑛,ℎ the generation of plant 𝑛 in hour ℎ, 𝑆𝑀𝑃𝐷𝐴ℎ the system
marginal price of the DAM in hour ℎ, 𝑉𝐶ℎ the variable cost of plant 𝑛, 𝑇𝑜𝑡𝐺𝑒𝑛𝐷𝐴𝑛 the
total generation of power plant 𝑛 in the DA market, then:
∑ℎ 𝐺𝑒𝑛𝐷𝐴𝑛,ℎ ∙ (𝑆𝑀𝑃𝐷𝐴ℎ − 𝑉𝐶𝑛 )
𝑂𝐶𝑛 =
𝑇𝑜𝑡𝐺𝑒𝑛𝐷𝐴𝑛
All hydro capacities (reservoirs) are assumed to bid in the UC_Simul. This is in
contrast to the DAM simulation, where only the non-mandatory part of lakes is
bidding, while the hydro generation that is assumed to be must-run production (due
to excess water and other uses of water) is excluded from bidding. In this way, the
simulation with the UC captures the role of hydro capacities for balancing purposes
(thus capturing their role in the Intra-day and balancing markets) and their
contribution to ancillary services.
Priority dispatching and demand response (if eligible to participate in the Intra-day
and balancing markets in the user-defined options) are incorporated in the UC
simulations the same way as they are incorporated in the DAM_Simul. The UC_Simul
optimization of real-time unit commitment reduces deviations from the DAM solution
as much as possible. For this reason, the model assumes penalty factors for the
deviations from the DAM schedule (re-dispatching), which are taken into account in
the objective function of the optimisation. Hence, re-dispatching costs occur for units
which operate according to the UC at a different level compared to the DAM schedule.
The deviations of the units’ dispatching schedule between the UC model and the DAM
model are financially settled in the simulations with the Intra-Day and Balancing
market simulator.

L.15.l. Intra-Day and Balancing simulator (IDB_Simul)

The Intra-day and Balancing Simulator (IDB_Simul) simulates a stylised hourly
market for the deviations that occur between the DAM_Simul solution and the
UC_Simul solution. The deviations are the result of the assumed errors in the day-
ahead forecast on load, RES generation, availability of plants, etc. which are generated
using the Random Events Generator. The simulator performs a single market clearing
of the deviations, i.e. it does not simulate continuous intra-day markets. It runs
simultaneously for all hours in every typical day, and determines an SMP price for
deviations, the financial settlement of deviations and a revised schedule for operation
of units as well as interconnectors.
Before running the IDM_Simul, for every hour, the nodes (countries) are categorized
in regions based on a specific coupling criterion; adjacent nodes are coupled in a
specific hour if and only if they share the same SMP for that hour.
Comparing the DAM and UC solutions, deviations in energy offers occur due to the
consideration of technical constraints of plants in the UC model and due to other
deviations in demand and renewables generation as adopted in the experiments with
the Random Event Generator. The IDB_Simul, before settling financially these
deviations, it further adjusts the unit commitment schedule of plants according to a
set of rules, which determine which resources are eligible to bid in the IDM to meet
the deviations. The bids are different for upward and for downward deviations of
power supplied by the eligible resources.
For each coupled region and for each hour, all the dispatchable power plants that have
altered their generation from the DAM solution to the UC solution opposite to the
direction of demand deviation (sum of the demand deviations of the countries in the
coupled region) are grouped separately. This group is further split into two
independent subgroups, one for every direction of the demand deviation. If demand in
the day-ahead simulation is lower than the one in UC and the generation of the unit is
higher in the day-ahead simulation than the generation in UC, then this plant is not
allowed to offer energy for upward deviations. If the reverse is true, then the power
plant is not allowed to offer for downward deviations. The logic behind this, is that
these plants are not load-following in the UC solution due to technical reasons, and
thus should not be able to contribute in covering intra-day deviation. Hence, the
deviations between DA and UC solutions in the generation volumes of the plants
belonging to these two groups have to be met by the rest of the conventional plants.
It remains to define the supply quantities each power plant can provide to IDM. The
majority of conventional power plants can participate in the IDM (including demand
response), with the exception of capacities that have been scheduled to participate in
the reserve and ancillary services market according to the UC solution, and the
capacities that are not load following as described in the paragraph above. When
deviations in demand are upward eligible power plants can potentially offer the entire
remaining capacity above the level that has been scheduled in DAM, minus any
amount qualified for upward reserve procurement. Equivalently, when deviations in
demand are downward they can potentially offer the entire remaining capacity below
the level that has been scheduled in DAM, minus their minimum generation level and
any amount qualified for downward reserves. Hydro generators in particular, can
offer energy only up to the maximum difference between DAM and UC solution, either
upwards or downwards.
Units which were not dispatched in the DAM solution are allowed to start-up during
the optimisation of IDB_Simul. Along the same lines, units which were dispatched in
the DAM solution are allowed to shut-down. However, this only applies to flexible
capacities (having ramping rates above a certain threshold value, as well as minimum
up time lower than a threshold value, both defined by the user), while inflexible
power plants are not allowed to either shut-down or start-up. None of the plants can
offer energy which violates their ramping rates.
Resources that are ultimately eligible to participate in the IDM compete each other
with their bidding offers. Energy offers for upward deviations are priced equal to the
marginal cost of the unit plus a scarcity mark-up in case there is shortage on
resources. Scarcity bidding methodology is the same as applied in the DAM_Simul.
Energy offer prices for downward deviations are driven by variable and fixed
operation and maintenance costs of each unit. Scarcity bidding applies here as well.
Shut-down or start-up decisions incur additional costs, as they do in UC model.
Bidding prices and ultimately remuneration of resources depend on the assumed
market liquidity, which varies across countries and across options. In countries and
options that no IDM market is assumed, generators receive administratively set prices
to cover for the deviations. In cases of illiquid markets, bidding is the same as the DAM
bidding. In cases of liquid markets, bidding is defined using DAM bids as a basis, plus a
mark-up reflecting scarcity in the market for deviations. Bids differ for upward and
downward deviations. Energy offers for upward deviations are priced equal to the
marginal cost of the unit plus a scarcity mark-up in case there is shortage on
resources. Energy offer prices for downward deviations are based on the variable and
fixed operation and maintenance costs of each unit, with scarcity mark-up applying as
well.
Network modelling in the context of ID market follows the same approach as with the
DAM_Simul. New power injections that occur during the optimization with the
IDM_Simul take into account congestion of transmission lines according to the UC
solution, in the sense that only residual capacity (beyond the schedule of the UC) of
interconnectors is eligible to participate in the IDM.

L.15.m. Reserve and Ancillary Services market or procurement simulator

(RAS_Simul)
In the simulation with the UC_Simul, binding of capacities for reserves has already
been determined. This amount of capacities is assumed to not participate in the IDM.
However, as the IDM determines an updated schedule of unit commitment compared
to the UC solution, it is probable that some capacities that were offering energy
according to the UC schedule offer less (or even shut-down) according to the IDM
schedule. Therefore, it is probable that there are additional capacities available to
participate in the reserve and ancillary services market compared to the UC solution.
It is assumed that only gas turbines are eligible for this purpose, due to high ramping
rates and short response times.
Thus, the RAS_Simul runs to re-settle financially the reserve and ancillary services
market taking into account the updated unit commitment schedule from the IDM. The
RAS_Simul uses the same demand for reserves as the UC model. Four types of reserves
have been considered:
1. Frequency Containment Reserve (primary reserve)
2. Automatic Frequency Restoration Reserve (secondary reserve –
Automation Generation Control)
3. Manual Frequency Restoration Reserve (spinning tertiary reserve)
4. Replacement Reserve (non-spinning tertiary reserve)
The following market function cases can be defined by the model user:
1. Procurement based on TSO contracts with specific plants, at defined prices.
2. Plants bidding for the reserve obligations, but receiving remuneration
based on administratively set prices
3. Competitive market for reserves with economic offers (bids) for prices and
quantities.
The RAS_Simul applies the same bidding for reserves as the UC model. Eligible
resources to participate in the market/procurement are differentiated according to
user-defined options, and can be all units which have been opted to participate in the
ID schedule, including demand response. In case participation of RES is allowed in the
ID market, they are only eligible to bid for downward reserve.
Demand response is allowed to participate in the RAS market (depending on user-
defined option) and is incorporated the same way as in the DAM_Simul.
 Resources available cross-border can also participate (differently constrained by
policy option) in the reserve market, subject to limitation from availability of
interconnection capacity, which is the capacity remaining after the schedule of the
IDM. Resources not scheduled in the IDM can submit bids to market for reserve, but
only for tertiary reserve.
The four reserve markets or procurements are inter-related because of technical
restrictions of the plants, and therefore run simultaneously.

L.15.n. Mathematical Illustration of PRIMES-IEM model

DAY-AHEAD MARKET SIMULATOR (DAM_SIMUL)
Known Parameters and Functions Unknown Variables
𝑑𝑖,ℎ Inverse demand function 𝑄𝑖,ℎ Consumption of electricity
𝑏𝑖,𝑛,ℎ Price bidding function 𝑃𝑖,ℎ System Marginal Price
Commitment schedule of power
𝑞𝑖,𝑛,ℎ Power quantities in priority dispatch 𝑞𝑖,𝑛,ℎ
plants
𝐾𝑖,𝑛,ℎ Power plant capacities 𝑈𝑝𝑖,𝑛,𝑎,ℎ Supply of upward ancillary service
Supply of downward ancillary
𝑀𝑖,𝑛,ℎ Technical minimum operation of a plant 𝐷𝑛𝑖,𝑛,𝑎,ℎ
service
𝑅𝑖,𝑛 Ramping capability of plant 𝑢𝑖,𝑛,ℎ Operating status of a plant (binary)
𝑀𝑢𝑝𝑖,𝑛,ℎ Minimum up time of a plant 𝑠𝑑𝑖,𝑛,ℎ Shut down of a plant (binary)
𝑀𝑑𝑛𝑖,𝑛,ℎ Minimum down time of a plant 𝑠𝑢𝑖,𝑛,ℎ Start-up of a plant (binary)
Inflows minus Outflows in a node of
𝑓𝑢𝑝𝑖,𝑛,𝑎,ℎ Price bidding for upward ancillary services 𝜎𝑖,ℎ
the network
𝑓𝑑𝑛𝑖,𝑛,𝑎,ℎ Price bidding for downward ancillary services 𝜃𝑖,ℎ Voltage phase angles at a node
Flows over interconnectors
𝑛𝑒𝑡𝑖,𝑘 Network topology matrix 𝑓𝑘,ℎ
(positive or negative)
𝜔𝑘,𝑘𝑘 Matrix of line admittances Sets
Nodes of the network (one or many
𝑇𝑘 Capacity of interconnectors 𝑖 𝑜𝑟 𝑖𝑖
per country)
𝑁𝑇𝐶𝑖,𝑖𝑖 Net Transfer Capacity between two nodes ℎ 𝑜𝑟 ℎℎ Time intervals (hours) in a year
𝑛 Power plants
𝑎 Reserve types
𝑘 𝑜𝑟 𝑘𝑘 Interconnectors
Equations for Day Ahead Market Simulator (energy only market)
𝑄𝑖,ℎ
−1
𝑧ℎ = ∑ (∫ 𝑑𝑖,ℎ (𝑦𝑖,ℎ ) 𝑑𝑦𝑖,ℎ − 𝑃𝑖,ℎ ∑ 𝑏𝑖,𝑛,ℎ (𝑃𝑖,ℎ )) Social Surplus to maximize
𝑖 0 𝑛
𝑃𝑖,ℎ = 𝑑𝑖,ℎ (𝑄𝑖,ℎ ) Inverse Demand Function
Price bidding by plant as function of
𝐵𝑖,𝑛,ℎ = 𝑏𝑖,𝑛,ℎ (𝑞𝑖,𝑛,ℎ )
volume
Capacity constraints of power
𝑞𝑖,𝑛,ℎ ≤ 𝐾𝑖,𝑛,ℎ
plants
𝜎𝑖,ℎ = ∑ 𝜃𝑖,ℎ 𝑛𝑒𝑡𝑖,𝑘 ∑ 𝜔𝑘,𝑘𝑘 𝑛𝑒𝑡𝑘𝑘,𝑖 Inflows minus Outflows in a node of
𝑖 𝑘𝑘
the network
𝑓𝑘,ℎ = − ∑ 𝜃𝑖,ℎ ∑ 𝜔𝑘,𝑘𝑘 𝑛𝑒𝑡𝑘𝑘,𝑖 Flows over interconnectors
𝑖 𝑘𝑘
(positive or negative)
Physical capacity constraint for
|𝑓𝑘,ℎ | ≤ 𝑇𝑘
flows over interconnectors
Restriction of bilateral flows due to
| ∑ ∑ 𝑓𝑘,ℎ | ≤ 𝑁𝑇𝐶𝑖,𝑖𝑖
Net Transfer Capacity
𝑘⊆𝑛𝑒𝑡𝑖𝑖,𝑘 𝑘⊆𝑛𝑒𝑡𝑖,𝑘
−1
𝑑𝑖,ℎ (𝑃𝑖,ℎ ) − ∑ (𝑞𝑖,𝑛,ℎ + 𝑞𝑖,𝑛,ℎ ) = 𝜎𝑖,ℎ Balance of inflows and outflows in a
𝑛
node
Equations for Unit Commitment Simulator (or Day Ahead Market with co-optimization of reserves)
 𝑄𝑖,ℎ
𝑧ℎ = ∑ (∫ 𝑑𝑖,ℎ (𝑦𝑖,ℎ ) 𝑑𝑦𝑖,ℎ
𝑖 0
−1 −1
− 𝑃𝑖,ℎ ∑ 𝑏𝑖,𝑛,ℎ (𝑃𝑖,ℎ ) − ∑ 𝑃𝑢𝑝𝑖,𝑎,ℎ ∑ 𝑓𝑢𝑝𝑖,𝑛,𝑎,ℎ (𝑃𝑢𝑝𝑖,𝑎,ℎ ) Social Surplus to maximize
𝑛 𝑎 𝑛

−1
− ∑ 𝑃𝑑𝑛𝑖,𝑎,ℎ ∑ 𝑓𝑑𝑛𝑖,𝑛,𝑎,ℎ (𝑃𝑑𝑛𝑖,𝑎,ℎ ))
𝑎 𝑛
𝑃𝑖,ℎ = 𝑑𝑖,ℎ (𝑄𝑖,ℎ ) Inverse Demand Function
∑ 𝑈𝑝𝑖,𝑛,𝑎,ℎ ≥ 𝐷𝑢𝑝𝑖,𝑎,ℎ Balance for upward ancillary services
𝑛
∑ 𝐷𝑛𝑖,𝑛,𝑎,ℎ ≥ 𝐷𝑑𝑛𝑖,𝑎,ℎ Balance for downward ancillary services
𝑛
𝐵𝑖,𝑛,ℎ = 𝑏𝑖,𝑛,ℎ (𝑞𝑖,𝑛,ℎ ) Price bidding by plant as function of volume
𝐵𝑢𝑝𝑖,𝑛,𝑎,ℎ = 𝑓𝑢𝑝𝑖,𝑛,𝑎,ℎ (𝑈𝑝𝑖,𝑛,𝑎,ℎ ) Bidding for upward ancillary services
𝐵𝑑𝑛𝑖,𝑛,𝑎,ℎ = 𝑓𝑑𝑛𝑖,𝑛,𝑎,ℎ (𝐷𝑛𝑖,𝑛,𝑎,ℎ ) Bidding for downward ancillary services
𝑞𝑖,𝑛,ℎ + ∑ 𝑈𝑝𝑖,𝑛,𝑎,ℎ ≤ 𝑢𝑖,𝑛,ℎ 𝐾𝑖,𝑛,ℎ Capacity constraints of power plants
𝑎
𝑞𝑖,𝑛,ℎ + ∑ 𝐷𝑛𝑖,𝑛,𝑎,ℎ ≥ 𝑢𝑖,𝑛,ℎ 𝑀𝑖,𝑛,ℎ Operation above technical minimum
𝑎
|𝑞𝑖,𝑛,ℎ − 𝑞𝑖,𝑛,ℎ−1 | ≤ 𝑢𝑖,𝑛,ℎ−1 𝑅𝑖,𝑛 + 𝑠𝑢𝑖,𝑛,ℎ 𝑅𝑖,𝑛 Ramping constraints
∑ 𝑠𝑑𝑖,𝑛,ℎℎ ≤ 1 − 𝑢𝑖,𝑛,ℎ Minimum down time constraint
ℎℎ∈[(ℎ−𝑀𝑑𝑛𝑖,𝑛,ℎ −1≤ℎℎ)∩(ℎℎ≤ℎ)]

∑ 𝑠𝑢𝑖,𝑛,ℎ ≤ 𝑢𝑖,𝑛,ℎ Minimum up time constraint

ℎℎ∈[(ℎ−𝑀𝑢𝑝𝑖,𝑛,ℎ +1≤ℎℎ)∩(ℎℎ≤ℎ)]
𝑢𝑖,𝑛,ℎ − 𝑢𝑖,𝑛,ℎ−1 = 𝑠𝑢𝑖,𝑛,ℎ − 𝑠𝑑𝑖,𝑛,ℎ Operation status constraint
𝑠𝑢𝑖,𝑛,ℎ + 𝑠𝑑𝑖,𝑛,ℎ ≤ 1 Shut or start constraint
𝜎𝑖,ℎ = ∑ 𝜃𝑖,ℎ 𝑛𝑒𝑡𝑖,𝑘 ∑ 𝜔𝑘,𝑘𝑘 𝑛𝑒𝑡𝑘𝑘,𝑖 Inflows minus Outflows in a node of the
𝑖 𝑘𝑘
network
𝑓𝑘,ℎ = − ∑ 𝜃𝑖,ℎ ∑ 𝜔𝑘,𝑘𝑘 𝑛𝑒𝑡𝑘𝑘,𝑖 Flows over interconnectors (positive or
𝑖 𝑘𝑘
negative)
Physical capacity constraint for flows over
|𝑓𝑘,ℎ | ≤ 𝑇𝑘
interconnectors
Restriction of bilateral flows due to Net
| ∑ ∑ 𝑓𝑘,ℎ | ≤ 𝑁𝑇𝐶𝑖,𝑖𝑖
Transfer Capacity
𝑘⊆𝑛𝑒𝑡𝑖𝑖,𝑘 𝑘⊆𝑛𝑒𝑡𝑖,𝑘
−1
𝑑𝑖,ℎ (𝑃𝑖,ℎ ) − ∑ (𝑞𝑖,𝑛,ℎ + 𝑞𝑖,𝑛,ℎ ) = 𝜎𝑖,ℎ Balance of inflows and outflows in a node
𝑛

INTRA-DAY AND BALANCING MARKETS SIMULATOR (IDB_SIMUL)

Known Parameters and Functions Unknown Variables
𝑢𝑝 𝑢𝑝 Upward balancing power output of power plants
𝐷𝑖,ℎ Upward deviations 𝑞𝑖,𝑛,ℎ
already opened
𝑑𝑜𝑤𝑛 𝑑𝑜𝑤𝑛 Downward balancing power output of power
𝐷𝑖,ℎ Downward deviations 𝑞𝑖,𝑛,ℎ
plants
𝑢𝑝 Price bidding function for upward 𝑜𝑝𝑒𝑛 Upward balancing power output of power plants
𝑏𝑖,𝑛,ℎ 𝑞𝑖,𝑛,ℎ
offers already opened
𝑑𝑜𝑤𝑛 Price bidding function for Deviation from DAM variable (binary, 1 if plant
𝑏𝑖,𝑛,ℎ 𝑝𝑜𝑠𝑖,𝑛,ℎ
downward offers committed in IDM and closed in DAM)
𝑑𝑎𝑚 Commitment schedule of power Deviation from DAM variable (binary, 1 if plant
𝑔̅𝑖,𝑛,ℎ 𝑛𝑒𝑔𝑖,𝑛,ℎ
plants from DAM committed in DAM and closed in IDM)
𝑢𝑐 Commitment schedule of power 𝑖𝑑 Operating status of a plant (binary) taken into
𝑔̅𝑖,𝑛,ℎ 𝑢𝑖,𝑛,ℎ
plants from UC account DAM schedule
Operating status of a plant 𝑖𝑑
𝑢̅𝑖,𝑛,ℎ 𝑠𝑑𝑖,𝑛,ℎ Shut down of a plant (binary)
(binary) from DAS and UC
𝑑𝑎𝑚 𝑖𝑑
𝐷𝑖,ℎ Demand from DAM 𝑠𝑢𝑖,𝑛,ℎ Start-up of a plant (binary)
𝑢𝑐
𝐷𝑖,ℎ Demand from UC
̅𝑑𝑎𝑚 Flows over interconnectors from
𝑓𝑘,ℎ
DAM
𝑑𝑎𝑚 Inflows minus Outflows in a node
𝜎̅𝑘,ℎ Sets
of the network from DAM
𝑠𝑢 Bidding for starting up a power
𝐶𝑖,𝑛,ℎ 𝑖𝑛𝑡(𝑛) Intermittent RES power plants
plant in IDM
 Power plants that are cannot offer up or down
𝑠𝑢 Bidding for shutting down a
𝐶𝑖,𝑛,ℎ deviation due to technical constraints or due to
𝑡𝑛(𝑛)
power plant in IDM
TSO instruction
Equations for Intra Day Ahead Market Simulator
𝑢𝑝 𝑢𝑝 𝑑𝑜𝑤𝑛 𝑑𝑜𝑤𝑛 𝑖𝑑 𝑠𝑢
𝑧ℎ = ∑ ∑ (𝐵𝑖,𝑛,ℎ 𝑞𝑖,𝑛,ℎ + 𝐵𝑖,𝑛,ℎ 𝑞𝑖,𝑛,ℎ + 𝑠𝑢𝑖,𝑛,ℎ 𝐶𝑖,𝑛,ℎ
𝑖 𝑛∉(𝑖𝑛𝑡∪𝑡𝑛) Cost of deviations to minimize
𝑖𝑑 𝑠𝑑
+ 𝑠𝑑𝑖,𝑛,ℎ 𝐶𝑖,𝑛,ℎ )

𝑢𝑝 𝑢𝑐 𝑢𝑐 𝑑𝑎𝑚 𝑑𝑎𝑚 𝑑𝑎𝑚

𝐷𝑖,ℎ = max( 𝐷𝑖,ℎ − 𝜎𝑖,ℎ − 𝐷𝑖,ℎ + 𝜎𝑖,ℎ + ∑ 𝑔̅𝑖,𝑛,ℎ
𝑛 ∈(𝑖𝑛𝑡∪𝑡𝑛)
Upward deviations
𝑢𝑐
− ∑ 𝑔̅𝑖,𝑛,ℎ , 0)
𝑛 ∈(𝑖𝑛𝑡∪𝑡𝑛)
𝑢𝑝 𝑑𝑎𝑚 𝑑𝑎𝑚 𝑢𝑐 𝑢𝑐 𝑢𝑐
𝐷𝑖,ℎ = max( 𝐷𝑖,ℎ − 𝜎𝑖,ℎ − 𝐷𝑖,ℎ + 𝜎𝑖,ℎ + ∑ 𝑔̅𝑖,𝑛,ℎ
𝑛 ∈(𝑖𝑛𝑡∪𝑡𝑛)
Downward deviations
𝑑𝑎𝑚
− ∑ 𝑔̅𝑖,𝑛,ℎ , 0)
𝑛 ∈(𝑖𝑛𝑡∪𝑡𝑛)
𝑢𝑝 𝑢𝑝 𝑢𝑝 Price bidding by plant as function of volume for
𝐵𝑖,𝑛,ℎ = 𝑏𝑖,𝑛,ℎ (𝑞𝑖,𝑛,ℎ )
upward offers
𝑑𝑜𝑤𝑛 𝑑𝑜𝑤𝑛 𝑑𝑜𝑤𝑛 Price bidding by plant as function of volume for
𝐵𝑖,𝑛,ℎ = 𝑏𝑖,𝑛,ℎ (𝑞𝑖,𝑛,ℎ )
upward offers
𝑢𝑝 Capacity constraint for upwards offers of power
𝑔̅𝑖,𝑛,ℎ + 𝑞𝑖,𝑛,ℎ ≤ 𝐾𝑖,𝑛,ℎ (1 − 𝑛𝑒𝑔𝑖,𝑛,ℎ )
plants
𝑑𝑜𝑤𝑛 Capacity constraint for downwards offers of
𝑞𝑖,𝑛,ℎ ≤ 𝑔̅𝑖,𝑛,ℎ (1 − 𝑛𝑒𝑔𝑖,𝑛,ℎ )
power plants
𝑜𝑝𝑒𝑛 Capacity constraint for upwards offers of power
𝑞𝑖,𝑛,ℎ ≤ 𝐾𝑖,𝑛,ℎ 𝑝𝑜𝑠𝑖,𝑛,ℎ
plants
𝑜𝑝𝑒𝑛 Technical minimum constraint for upwards
𝑞𝑖,𝑛,ℎ ≥ 𝑀𝑖,𝑛,ℎ 𝑝𝑜𝑠𝑖,𝑛,ℎ
offers of power plants
𝑢𝑝 𝑢𝑝
|𝑞𝑖,𝑛,ℎ − 𝑞𝑖,𝑛,ℎ−1 | ≤ 𝑅𝑖,𝑛 Ramping Constraint for upwards offers
𝑜𝑝𝑒𝑛 𝑜𝑝𝑒𝑛 𝑖𝑑
|𝑞𝑖,𝑛,ℎ − 𝑞𝑖,𝑛,ℎ−1 | ≤ 𝑝𝑜𝑠𝑖,𝑛,ℎ−1 𝑅𝑖,𝑛 + 𝑠𝑢𝑖,𝑛,ℎ 𝑅𝑖,𝑛 Ramping Constraint for upwards offers
𝑑𝑜𝑤𝑛 𝑑𝑜𝑤𝑛
|𝑞𝑖,𝑛,ℎ − 𝑞𝑖,𝑛,ℎ−1 | ≤ 𝑅𝑖,𝑛 Ramping Constraint for downwards offers
𝑖𝑑
𝑢𝑖,𝑛,ℎ = 𝑢̅𝑖,𝑛,ℎ + 𝑝𝑜𝑠𝑖,𝑛,ℎ − 𝑛𝑒𝑔𝑖,𝑛,ℎ Commitment Constraint
𝑖𝑑
𝑝𝑜𝑠𝑖,𝑛,ℎ − 𝑝𝑜𝑠𝑖,𝑛,ℎ−1 ≤ 𝑠𝑢𝑖,𝑛,ℎ Start-up constraint in IDM
𝑖𝑑
𝑛𝑒𝑔𝑖,𝑛,ℎ − 𝑛𝑒𝑔𝑖,𝑛,ℎ−1 ≤ 𝑠𝑑𝑖,𝑛,ℎ Shut-down constraint in IDM
𝑖𝑑 𝑖𝑑
𝑠𝑢𝑖,𝑛,ℎ + 𝑠𝑑𝑖,𝑛,ℎ ≤1 Shut or start constraint for IDM
𝑝𝑜𝑠𝑖,𝑛,ℎ + 𝑛𝑒𝑔𝑖,𝑛,ℎ ≤ 1 Shut or start deviation constraint
∑ 𝑛𝑒𝑔𝑖,𝑛,ℎℎ ≤ 1 Minimum down time constraint
ℎℎ∈[(ℎ−𝑀𝑑𝑛𝑖,𝑛,ℎ −1≤ℎℎ)∩(ℎℎ≤ℎ)]

∑ 𝑝𝑜𝑠𝑖,𝑛,ℎ ≤ 1 Minimum up time constraint

ℎℎ∈[(ℎ−𝑀𝑢𝑝𝑖,𝑛,ℎ +1≤ℎℎ)∩(ℎℎ≤ℎ)]
𝑑𝑎𝑚
𝜎𝑖,ℎ + 𝜎̅𝑘,ℎ = ∑ 𝜃𝑖,ℎ 𝑛𝑒𝑡𝑖,𝑘 ∑ 𝜔𝑘,𝑘𝑘 𝑛𝑒𝑡𝑘𝑘,𝑖 Inflows minus Outflows in a node of the network
𝑖 𝑘𝑘
̅𝑑𝑎𝑚 = − ∑ 𝜃𝑖,ℎ ∑ 𝜔𝑘,𝑘𝑘 𝑛𝑒𝑡𝑘𝑘,𝑖
𝑓𝑘,ℎ + 𝑓𝑘,ℎ Flows over interconnectors (positive or negative)
𝑖 𝑘𝑘
̅𝑑𝑎𝑚 | ≤ 𝑇𝑘 Physical capacity constraint for flows over
|𝑓𝑘,ℎ + 𝑓𝑘,ℎ
interconnectors

̅𝑑𝑎𝑚 | ≤ 𝑁𝑇𝐶𝑖,𝑖𝑖 Restriction of bilateral flows due to Net Transfer

| ∑ ∑ 𝑓𝑘,ℎ + 𝑓𝑘,ℎ
Capacity
𝑘⊆𝑛𝑒𝑡𝑖𝑖,𝑘 𝑘⊆𝑛𝑒𝑡𝑖,𝑘
𝑢𝑝 𝑑𝑜𝑤𝑛 𝑢𝑝 𝑜𝑝𝑒𝑛𝑑𝑜𝑤𝑛
𝐷𝑖,ℎ − 𝐷𝑖,ℎ − ∑ (𝑞𝑖,𝑛,ℎ + 𝑞𝑖,𝑛,ℎ − 𝑞𝑖,𝑛,ℎ ) = 𝜎𝑖,ℎ Balance of inflows and outflows in a node
𝑛∉(𝑖𝑛𝑡∪𝑡𝑛)

RESERVE AND ANCILLARY SERVICES MARKET SIMULATOR (RAS_SIMUL)

Known Parameters and Functions Unknown Variables
𝑖𝑑 Commitment schedule of power plants Contribution of flows to reserve and ancillary
𝑔̅𝑖,𝑛,ℎ 𝑐𝑖,ℎ
after IDM service market

𝑢𝑝 Upper limit of contribution of flows to

𝑐𝑖,ℎ
RAS
̅𝑖𝑑 Flows over interconnectors from DAM
𝑓𝑘,ℎ
and IDM
 𝑖𝑑 Inflows minus Outflows in a node of the
𝜎̅𝑘,ℎ
network from DAM and ID
Equations for Reserve and Ancillary Services Simulator
𝑧ℎ = ∑ ∑ ∑(𝐵𝑢𝑝𝑖,𝑛,𝑎,ℎ 𝑈𝑝𝑖,𝑛,𝑎,ℎ + 𝐵𝑑𝑛𝑖,𝑛,𝑎,ℎ 𝐷𝑛𝑖,𝑛,𝑎,ℎ ) Cost of ancillary services to minimize
𝑖 𝑎 𝑛
𝐵𝑢𝑝𝑖,𝑛,𝑎,ℎ = 𝑓𝑢𝑝𝑖,𝑛,𝑎,ℎ (𝑈𝑝𝑖,𝑛,𝑎,ℎ ) Bidding for upward ancillary services
𝐵𝑑𝑛𝑖,𝑛,𝑎,ℎ = 𝑓𝑑𝑛𝑖,𝑛,𝑎,ℎ (𝐷𝑛𝑖,𝑛,𝑎,ℎ ) Bidding for downward ancillary services
𝑖𝑑
𝑔̅𝑖,𝑛,ℎ 𝑖𝑑
+ ∑ 𝑈𝑝𝑖,𝑛,𝑎,ℎ ≤ 𝑢𝑖,𝑛,ℎ 𝐾𝑖,𝑛,ℎ Upper bound of contribution to upward
𝑎
ancillary service constraint for power plants
𝑖𝑑
𝑔̅𝑖,𝑛,ℎ 𝑖𝑑
+ ∑ 𝐷𝑛𝑖,𝑛,𝑎,ℎ ≥ 𝑢𝑖,𝑛,ℎ 𝑀𝑖,𝑛,ℎ Upper bound of contribution to downward
𝑎
ancillary service constraint for power plants
𝑢𝑝 Upper bound of contribution to upward
𝑐𝑖,ℎ ≤ 𝑐𝑖,ℎ
ancillary service constraint for x-border flows
𝑖𝑑
∑ 𝑈𝑝𝑖,𝑛,𝑎,ℎ ≥ 𝐷𝑢𝑝𝑖,𝑎,ℎ + 𝑐𝑖,ℎ 𝜎̅𝑘,ℎ Balance for upward ancillary services
𝑛
∑ 𝐷𝑛𝑖,𝑛,𝑎,ℎ ≥ 𝐷𝑑𝑛𝑖,𝑎,ℎ Balance for downward ancillary services
𝑛

L.15.o. Methodology of the Modelling of Power Generation Investment

under Uncertainty (optional)
The version of the power sector model of PRIMES which simulates oligopoly
competition includes an option of valuing investment in power generation under
uncertainty.
The simulation of oligopoly competition computes electricity market prices which
define a stream of revenues for existing, or candidate for investment, plant. The
revenues seen from the perspective of an investor or of a plant owner, support the
decision about going forward with the investment or not. Similarly, it supports the
decision for a plant owner about whether to maintain the plant in operation, or
instead retiring the plant from the market. This “investment evaluation” process is
simulated using the Investment Evaluation model.
The methodology followed to evaluate the value of power plants (existing or
candidate for investment) considers uncertainty about the market in the future and
heterogeneity of decision makers regarding the hurdle rates.
In order to account for uncertain market conditions, the approach introduces three
random variables, namely ETS prices, gas prices and RES development. To introduce
randomness around these factors we assume that each follows a Brownian motion
which has been applied to the whole time series trajectory. Moreover, we take into
consideration the interdependencies of the three random variables. The trajectory of
the random variables in time is a random process which is the outcome of a 3-
dimensional stochastic differential equation, having a pre-determined mean (using
PRIMES projections for this purpose) and a covariance matrix defined so as to reflect
the relationship between the three random variables.
ETS prices and gas prices are positively correlated, as a future of higher gas prices
would imply a future where coal-based and other high-emitting generation is more
competitive, and would therefore require higher ETS prices to achieve the EUCO
emissions targets. ETS prices and RES development are negatively correlated, as the
larger the development of RES the lower the required level of ETS prices for
maintaining the EUCO emissions targets. Finally, gas prices and RES development are
positively correlated as higher gas prices render RES more competitive.
With this basis, the model applies a Monte Carlo simulation technique to generate a
large sample of scenarios, each being representative of a particular “event” of ETS
prices, gas prices and RES development. After reducing the number of random
scenarios-events of our sample following a scenario reduction technique, the PRIMES
Oligopoly model for each scenario-event computes the stream of revenues of each
plant.
These streams of revenues are then used to calculate the plant specific present values
(PVs) for each scenario-event. In particular, the PV is calculated as present value of
revenues from the wholesale, balancing, ancillary services and CM markets, minus
variable, fuel and O&M costs.
The approach, instead of considering specific rates of return for the PV calculations,
assumes that plant owners are heterogeneous, and therefore considers a range of
hurdle rates (desired rates of return) which are assumed to be normally distributed.
The mean of the hurdle rates distribution is different for each Member State,
reflecting the varying financing conditions in each. The user can vary both the mean
value and the standard deviation of the probability distribution for the hurdle rates.
The idea is that the distribution of hurdle rates depends on the competition context
and the perceived certainty surrounding future revenues. This is assumed in order to
reflect that under different market conditions the behaviour of plant owners would
alter, and they would be willing to accept different rates of return (hurdle rates) for
undertaking a project. In particular, when future revenues are considered to be more
certain, investors are willing to accept lower rates of return (hurdle rates) to
undertake a project, and vice versa. Similarly, in conditions of more intense
competition investors tend to adopt lower hurdle rates, as otherwise they risk staying
out of the market. These changes of the hurdle rates are purely behavioural, and very
hard to quantify. However, this behaviour is key in the simulation and comparative
analysis of the different options of this study and they need to be reflected to the
extent possible. Following this logic, the investment evaluation process takes into
account assumptions whether the revenues come solely from the wholesale market,
which are uncertain as in a spot market, or also from capacity mechanisms or other
similar forms of securities for the capital costs, which are more certain than in spot
markets. Depending on the origin of revenues, the model applies for example lower
hurdle rates and possibly lower standard deviation of the hurdle rates in the case of
capacity mechanisms compared to the cases without.
The mathematical illustration follows. We may denote:

 𝑠: a scenario-event, with probability of occurrence 𝜋𝑠

 𝜌: the various types of decision makers, each applying a different hurdle rate,
with frequency 𝜋𝜌
 𝑖: a power plant
Then, the process described so far leads for every plant, old or new, to a collection of
present values 𝑃𝑉𝜌,𝑠,𝑖 , each with probability 𝜋𝑠 ∙ 𝜋𝜌 .

The next step of the investment evaluation process is to compare this set of PVs with a
benchmark, in order to assess how each project (plant) performs, and based on this
assessment decide upon its viability. In other words, the evaluation needs to specify a
probability that an investment is realized (or that an old project continues to operate)
as a function of its performance. Two measurements need therefore to be defined: a) a
measurement of the performance of each plant, and b) a probability function of
deciding positively on investing on (or continue operation of) each plant, based on its
performance.
New plants are considered to perform adequately if revenues minus costs are
sufficiently high to counterbalance investment expenditures. In case the PV of
revenues minus costs is negative, then definitely the investment should not be
realized, or in other words, the probability that the investment is realized should be
zero. If the PV is positive, then the probability that the investment is realized should
be getting higher with higher values of PV.
For old plants, the approach is different; an old plant is considered to perform well if
revenues are sufficiently high to cover for fixed and O&M costs, i.e. when the PV is
positive. A positive PV for old plants should imply that the probability that the plant
retires prematurely is zero. But even if revenues do not suffice to cover for these costs
(PV is negative), still the operation of the plant is of some worth to the decision maker
due to its salvage value (i.e. the non-amortized investment cost, calculated specifically
for each plant considering its remaining lifetime and investment cost), which is lost in
case of premature retirement. Therefore, the probability that the plant retires should
be increasing as the negative present value is increasing in absolute terms.

Following the above logic, we developed a “Performance ratio”, denoted 𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖 . The
Performance ratio is calculated differently for old and new plants. For new plants
(denoted 𝑖𝑛𝑒𝑤), the Performance ratio is the ratio of the present value of revenues
over the cost of investment (𝛪𝑖𝑛𝑒𝑤 ):
𝑃𝑉𝜌,𝑠,𝑖𝑛𝑒𝑤
𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑛𝑒𝑤 = ⁄𝛪
𝑖𝑛𝑒𝑤

For old plants (denoted 𝑖𝑜𝑙𝑑), the Performance ratio takes an inverse form and is the
ratio of minus the salvage value of the plant (𝑆𝑖𝑜𝑙𝑑 ) over the present value of revenues:

−|𝑆𝑖𝑜𝑙𝑑 |
𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑜𝑙𝑑 = ⁄𝑃𝑉
𝜌,𝑠,𝑖𝑜𝑙𝑑

Ultimately, the Performance ratio is used to derive a probability that a new

investment project is undertaken or that an old project is continuing to operate. This
probability function, referred to as probability of adequate performance and denoted
𝜋𝜌,𝑠,𝑖 , has been specified:

 For new plants

𝜋𝜌,𝑠,𝑖𝑛𝑒𝑤 = 𝑓(𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑛𝑒𝑤 ), 𝑓𝑜𝑟 𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑛𝑒𝑤 ≥ 0

𝜋𝜌,𝑠,𝑖𝑛𝑒𝑤 = 0, 𝑓𝑜𝑟 𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑛𝑒𝑤 < 0

𝑎
ln (1 + 𝑒 −𝑎(𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑛𝑒𝑤 ) )
𝑓(𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑛𝑒𝑤 ) = 1 − , 𝑓𝑜𝑟 𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑛𝑒𝑤 ≥ 0
ln(2)

 For old plants

𝜋𝜌,𝑠,𝑖𝑜𝑙𝑑 = 1 − 𝑓(𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑜𝑙𝑑 ), 𝑓𝑜𝑟 𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑜𝑙𝑑 ≥ 0

𝜋𝜌,𝑠,𝑖𝑜𝑙𝑑 = 1, 𝑓𝑜𝑟 𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑜𝑙𝑑 < 0

𝑎
−𝑎(𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑜𝑙𝑑 )
ln (1 + 𝑒 )
𝑓(𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑜𝑙𝑑 ) = , 𝑓𝑜𝑟 𝑃𝑒𝑟𝑓𝜌,𝑠,𝑖𝑜𝑙𝑑 ≥ 0
ln(2)
For new plants, the function is defined so as to ensure that a negative performance
ratio (resulting from a negative PV) leads to zero probability of an investment being
realized. For positive performance ratios, the probability takes a form that resembles
 an S curve, and of course ranges between 0 and 1. For small values of the performance
ratio, the probability is still very close to zero, as a decision maker would not be
willing to decide positively on an investment where the PV
of revenues minus costs would be considerably lower than
investment expenditure. This reflects the level of risk
Probability of adequate performance of plant i (πρ,s,i)

aversion assumed.

For old plants, the 𝜋𝜌,𝑠,𝑖 function ensures that those with
negative performance ratio (i.e. positive PV) are definitely
continuing operation. Plants with positive performance
ratio (negative PV), will continue operation with
probability varying from zero to one. If the absolute value
of PV is close to zero, then the probability that the plant
will survive is still quite large. However, as the PV takes
higher absolute values (implying that revenues are
becoming less and less compared to the costs) and the
performance ratio is getting close to one, then the
probability that the operation of the plant will continue
diminishes.
Finally, the process calculates the probability of survival of
In

Performance ratio of plant I (Perfρ,s,i)

each plant denoted as 𝑃𝑟𝑜𝑏𝑆𝑢𝑟𝑣𝑖 by multiplying the
probability of each scenario-event, times the frequency of
i: power plant the decision maker types, times the probability of
ρ: type of investor
adequate performance, and summing over the whole
range of possibilities:
s: scenario event

𝑃𝑟𝑜𝑏𝑆𝑢𝑟𝑣𝑖 = ∑ ∑ 𝜋𝜌,𝑠,𝑖 𝜋𝑠 𝜋𝜌
𝑠 𝜌

This probability of survival is multiplied with the capacity of each plant in a scenario
and yields with an updated capacity level. Next, these reduced capacities are used to
run again the PRIMES Oligopoly model to compute power generation, revenues and
total costs for consumers.
 M. The PRIMES Gas Supply Model
M.1. Scope of the model
The PRIMES energy system model includes a detailed gas supply module that
provides projections for gas imports by country of origin, by transport mean (LNG,
pipeline) and route as well as wholesale gas prices. The gas model studies the
relationships between gas resources, gas infrastructure and the degree of competition
in gas markets over the Eurasian and MENA area and evaluates their impacts on gas
prices paid by gas consumers in the EU Member-States.
The gas model is a dynamic market competition model, which covers the entire
Eurasian/MENA areas and the global LNG market. It presents in detail the gas
infrastructure, present and future, as well as the different “agents” that participate in
the market. The agents compete for access to gas infrastructure and for gas supply to
customers, the latter being responsive to gas prices. The model considers the
oligopolistic structure of the gas market, which includes market imperfections15 and
can accommodate different assumptions about the degree of competition and the
integration of the EU gas internal market.
The gas supply module uses as input the gas demand projections, available from the
end-use sectors for demand (twelve industrial sectors, transport, residential, services
and agriculture) and electricity generators of the PRIMES model. The model
determines the equilibrium by finding the prices such that the quantity producers find
best to supply matches the quantity consumers wish to use. Thus, the flow of gas over
the entire gas network, the economic decisions of the agents and the market prices
are endogenous and are computed dynamically. The module operates on an inter-
temporal basis from 2000 to 2030 and produces results by five year period.

M.2. Gas infrastructure

The gas module represents in detail the present and future gas infrastructure of each
EU Member State, other European countries and the gas producing and consuming
countries of the Eurasian/MENA area, including Russia, Ukraine, Belarus, the Caspian
countries, Middle East, Persian Gulf and North African countries. The model also
represents the supply possibilities of LNG worldwide and the global demand and
trade of LNG. The infrastructure types include: gas pipelines (represented as a
network), gas storage, LNG terminals, gas production and gas liquefaction.
The interregional flows of gas are simulated based on a gas transport network
consisting of high-pressure gas pipelines and ship routes for LNG. A detailed
Natural Gas is of representation of the physical natural gas pipeline system is used to represent the
great importance in current and possible interregional transfers under engineering constraints, allowing
the transition
towards a clean the moving of gas from producers to end-users. Simultaneously with physical flows,
and renewable the model projects the commercial transactions between suppliers and customers,
energy system.
Secure and which extend beyond neighbouring countries, involve transit routes and gas swaps to
affordable gas allow their implementation. Gas traders (arbitragers) are also included as well as gas
supply in the EU is transport system operators. Each country is assumed to have a single Transport
of critical
importance. System Operator (TSO), which manages flows coming into and out of the region. Each
TSO represents a transhipment node in the gas supply module. Arcs, defined as routes
carrying gas flowing between TSOs, connect the nodes. Each arc corresponds to an
aggregation of the pipelines between neighbouring countries.

15In technical terms the model solves a Nash-Cournot oligopoly game with conjectural variations to find
imperfect market equilibrium.
Arcs are also established from gas producers (fields) to transhipment nodes (TSOs)
and to gas liquefaction plants. Only one gas producer and one LNG producer, if
applicable, are considered by country, whereby the major gas (fields) and/or the
liquefaction plants are represented for each country. In addition, the gasification
plants (one by one) and the storage facilities (aggregated by country) are represented.
The supply from each country is directly available to only one transhipment node. If
the supply is made available to other countries (at an adjoining transhipment node), it
needs first to pass through a transhipment node.
Detailed cost data (capital and variable operating) are associated with each type of
gas infrastructure and so gas transportation costs, including LNG ship costs are
calculated as a function of distances.
The final consumer gas prices are explicitly computed and reflect costs but also
include market-related and depletion-related rents.
Gas supply and demand are balanced on a daily basis. A few typical days are
represented per country. Variability of gas demand is determined bottom-up from the
gas load profiles of various gas uses by sector as these are projected by the rest of the
PRIMES model.
Among the supply sources gas storage is represented: storage inputs and outputs
connected to each transhipment node to represent net storage withdrawals in the
country as needed to manage gas balancing at peak times. During the off-peak period,
net storage injections are calculated to establish gas storage balance over a year time
period. The gasification plants are modelled also as storage facilities having more
limited capacities than the underground storage facilities.
Third party access is assumed for gas infrastructures and regulated tariffs are applied,
which are determined by the model by mimicking current practices based on
regulated asset basis pricing methods. Exogenous parameters may be used to reflect
different regulatory policies for pricing, access and use of gas infrastructure. Also
long-term contracts are included as constraints between suppliers and customers.
Duration and terms of existing (in the beginning of projections) long term contracts
are exogenous. Upper and lower variability margins of flows over pipelines, reflecting
physical and/or contractual limitations, are represented through exogenous
parameters and constraints.

M.3. Modelling of competition

Gas producers and gas suppliers (traders) are considered as separate companies. A
gas supplier and/or trader is assumed to have access to a limited number of gas
production (or LNG) nodes and to a subset of gas demand nodes. This can vary by
scenario to reflect different degrees of competition intensity and market integration
in the EU. The traders are assumed to operate as financial brokers to profit from gas
price differences between country-specific demand nodes. It is also assumed that each
country-specific TSO operates a gas pool market (a hub) per country, which seeks to
maximise consumer and producer surplus under imperfect competition among
suppliers and price-elastic demand.
The TSOs operate as regulated monopolies and seek a regulated maximisation of
profits from balancing demand and supply at each node. The TSOs perform a daily
balancing of gas demand and supply, only in terms of “mass” of gas.
 Lique-
Gas
faction
Field

TSO
LNG
TSO
TSO Terminal
TSO
Storage Demand
Country

Pipeline capacities and investments are exogenous. Volume dependent curves are
specified for computing tariffs for transportation between transhipment nodes (e.g.
Member States and neighbouring – transit - countries). The tariff curves extend
beyond current pipeline capacity levels and relate incremental capacity to
corresponding estimated rates. The TSOs charge tariffs and apply mark-ups for
transportation service.
Gas producers (fields) seek the maximization of rents from inter-temporal
management of their exhaustible resources, selling to the pool managed by the TSO,
while the LNG producers, gas storage and LNG storage operators maximize the
profits from exploiting the storage facility. Gas production costs and potential rents
are represented by cost-supply curves with increasing slope, constrained by resource
potential. Use of gas facilities entails variable (nonlinear) and fixed costs.
Gas field reserve amounts are specified exogenously in the base year and based on the
model extraction follows a net depletion profile while reserves includes increases in
amounts due to exploration and recovery evolving in the future. Liquefaction, storage
and LNG gasification capacities and investment are exogenous. The dates of
commissioning of new infrastructures are exogenous.
Suppliers
Traders
Consumers
TSO Residential
Producers Services
TSO
(Gas fields Industry
and LNG) TSO
Power sector
TSO Other
sectors

Storage and gasification

The link between the TSOs and the gas consumers (residential sector, services sectors,
agriculture sector, transport, industrial sectors and electric generators) correspond to
gas flows, which implement the commercial gas flows between gas suppliers,
traders and customers. Gas suppliers and traders seek to maximize profits by
generating revenues from gas sales to consumers, while they incur costs by
purchasing gas from pools that are managed by TSOs.
The demand functions of gas consumers are price elastic. Demand detailed by gas
load segment and by sector is linked with the rest of PRIMES models.
The gas supply model determines oligopolistic market equilibrium over multiple
period years (up to 2050) and calculates the market prices of gas and LNG by country,
year and marginal system gas prices by load segment (typical days). The model
projects physical gas flows over the entire Eurasian gas infrastructure system,
calculates possible congestion for each gas facility (pipeline, LNG terminal, storage,
etc.) and evaluates financial balances (costs versus revenues and rates of use) by gas
infrastructure component. Thus, the model can support profitability analysis of new
gas infrastructure (new pipelines, LNG terminals, etc.).

M.4. Model usage

The gas supply model has been used for specific gas sector policy analyses:
• Congestion and profitability of new gas transport routes
• Alternative scenarios about development of new gas suppliers
• Impact of gas shortages for certain upstream suppliers
• Changes in the global market for LNG and their impacts
• Impact of reduced gas demand in the EU owing to energy efficiency and RES
• Impact of growth of domestic gas demand within major gas suppliers outside
Europe
• Impacts on the EU gas supply of growing gas demand in Asia
Summary of PRIMES Gas Supply Model

The Gas Supply model simulates an oligopoly market over multiple countries,
involving many actors (consumers, TSOs, traders and upstream producers)
• Consumers are price takers with demand being elastic with prices
• TSOs manage gas hubs and minimize cost of gas supply
• Traders maximise profits, perform arbitraging operations and are price
takers from upstream producers
• Upstream producers compete along a Nash-Cournot game (with
conjectural variations)
• The number of competitors acting on each node change over time to
reflect growing competition (long term trend towards a well-functioning
market)

Operations and flows are constrained by a physical system involving pipelines,

LNG terminals, gas storage facilities, liquefaction plants and gas producing wells.
The market clearing for pipeline gas is on a Eurasian scale, while for LNG the
coverage is global. Investment in gas infrastructure is exogenous. Characteristics
of gas companies are also exogenous.

The model simulates two layers of flows: physical gas flows and commercial
transactions
• A consumer on one node can be commercially supplied with gas produced
at a node without direct link with the consumption node (e.g. if gas swaps
implement the commercial transaction)
• Since the gas network constraints are binding, gas supply prices differ by
node (country)
• Price determination reflects marginal costs, an endogenous mark-up and
fixed costs that recover cost of infrastructure
• Upstream producers tariff gas according to a gas cost function inclusive of
gas field exhaustion rents (Hoteling’s rule)
The model simulates gas balancing on a daily basis, considering load
characteristics of gas demand sectors and possibilities of storing gas and using
LNG
 N. Oil Products Supply Model and biofuel blending
The refinery sub-model of PRIMES is used to project domestic components of
petroleum product prices, refining activities and refinery capacity expansion,
including where appropriate technological change. Crude oil prices are exogenous to
PRIMES.
The petroleum supply market in the EU is modelled as one stylised refinery by
country involving distillation and cracking processing facilities which differ by
country and are projected to the future through endogenous investment. The generic
processing units are atmospheric distillation, vacuum distillation, coking, catalytic
cracking, hydrocracking, and visbreaking. Net imports of petroleum products by
European country are projected according to time trends and relative prices using a
simple reduced-form import-export graph representation.
The model projects changes in the structure of refinery activities (building and
composition of processing units and energy consumption) to produce an inter-
temporal least cost product mix that satisfies the demand projected by the rest of
PRIMES energy demand and supply modes. The activities are constrained by specific
technical constraints, which simulate operation of refining over a stylised graph,
which includes intermediate streams as flows between processing units. Capacity is
allowed to expand, under technical limitations. Investment schedules are developed
endogenously. The regulatory framework concerning product quality and types are
incorporated as simple additive factors that increase cost of production.
End-use petroleum product prices are formed as function of crude oil cost recovery
(which are exogenous) marginal costs of refining, plus transportation costs,
distribution costs and taxes. Marginal costs are allocated to product types based on
results of the refining optimisation.
An add-on modelling in petroleum supply model projects blending of oil products
with biofuels, for gasoline, diesel, kerosene and fuel oil. The modelling of blending
follows the perspective of product suppliers who are constrained by policy to reduce
average emission factors of the blended products or to meet blending regulations. In
doing so they define blending percentages to meet regulations, to respect technical
constraints concerning combustion of the blended product and to minimise cost of
blended products. The suppliers are price takers of biofuel components as these are
determined in the biomass supply model of PRIMES. For cost minimisation the
suppliers take into account carbon prices/values as price signals of emission
reduction policies. Because of the blending, prices of blending products and their
average emission factors are determined and are then used as inputs by other PRIMES
sub-models.

The oil refining The following figure illustrates the structure of the stylised refinery modelled in
industry in the EU PRIMES for each country.
is likely to face
declining demand
for oil product and
environmental
constraints,
including bio-fuel
blending
regulations.
 O. Rest of Energy Branch related to fossil fuels
Primary The PRIMES model database includes data on domestic potential fossil fuel resources
by country, covering crude oil, shale oil, natural gas, and solid fuels (coal and lignite).
production of The reserve data have the form of cost-quantity curves with increasing slopes.
fossil fuels Extraction activity by country and by fuel type is projected using reduced-form
equations. Drivers are demand for fuels (projected in the rest of the PRIMES model),
international prices of fossil fuels (used to evaluate profitability of domestic
extraction by comparing international prices to domestic costs based on the cost-
quantity curves), policy-driven parameters which promote domestic production of
fossil fuels by setting lower limits on production schedule or by subsidizing domestic
costs. The fossil fuel extraction model solves inter-temporarily. Projection of natural
gas extraction is coordinated with projections by the more detailed gas supply model.

Solid Fuel The model calculates in a simple way inputs and outputs of plants that convert solid
energy forms. The following are included: briquetting of coal and lignite; coke
Conversion production from coal; coke-oven-gas production; blast furnace gas production. Output
Plants from such plants depends on demand calculated by the rest of the PRIMES model.
Energy consumption, losses and emissions are projected using reduced form
equations, which take into account efficiency improvement possibilities and
environmental policies, including carbon prices.
The coke and derived gas activities are related to the existence and operation of blast
furnaces in a country.

Other The model includes representation of processes, which convert fossil fuels such as
coal liquefaction, oil gasification, production of gas works, and recovery of oil
conversion feedstock. Inputs and outputs are based on simple technical descriptions of the
processes processes. Activity depends on demand and on time trends, which are specific by
country. Gas works is a declining activity. The rest of the processes depend on
domestic primary production of fossil fuels. Energy consumption, losses and
Losses of oil emissions are calculated.
and gas Losses of oil pipelines, as well as of gas transportation and gas infrastructure are
transportation calculated using fixed loss ratios. The calculation is integrated in the gas supply sub-
model and distinguishes between different types of gas infrastructures.

Non Energy The model treats petrochemical consumption of fuels for energy and non-energy
purposes (raw material) within the sub-model treating the chemicals sector and its
Consumption sub-sector. Other fuels used for non-energy purposes (e.g. asphalt) are linked to
activity of the construction sector.

Energy The PRIMES model computes in detail energy consumption by fuel type and electricity
that are used by sectors producing or converting fossil fuels. Fuels are used by motor
Consumption drives and engines. Steam and electricity are used in specific energy uses. The list of
by Energy sub-sector of the energy branch is the following: Coal, Lignite extraction; Gas, Oil
extraction; Briquetting; Coke production; Gas works; Pipelines and compressors; LNG
Branch
terminals; Nuclear fuel and waste; Refineries; Energy in other Transport uses (port
cranes and airport vehicles).
 P. Projection of Energy Balances
PRIMES produces Excel reports containing projected energy balances by country
Projection of
following the detailed format and statistical conventions of Eurostat. The projection
Eurostat figures come from the various PRIMES sub-models and are fully balanced, in the sense
balances that they respect all balancing conditions of Eurostat methods, including balancing at:
gross inland consumption and supply, demand and supply at final energy
consumption, balance of transfer between products, consistency between inputs,
outputs, losses and energy consumption of each energy transformation activity, etc.
All figures are measured in ton of oil equivalent. Reporting of projected emissions is
also detailed as the energy balances.

Treatment of The projected energy balances have two distinct forms depending on the treatment of
fuels uses to produce steam by on-site industrial CHP units (the CHP plants which do
on-site not sell steam to other users):
industrial
a) as in Eurostat methods, i.e. by including these fuels in industrial energy
CHP consumption and by not showing the steam produced by the corresponding
on-site industrial CHP;
b) as in PRIMES, i.e. by including these fuels in transformation input tables (for
power/steam generation), by including the corresponding steam in
transformation output tables (for power/steam generation) and by showing
the steam amounts in final consumption of industry (by sector).
The data on on-site industrial CHP are produced during PRIMES model calibration
using data from Eurostat CHP surveys and other sources of information by industry
sector. The calibration model takes cares to produce consistent splits of CHP between
on site and distributed steam activities by sector, to match the more aggregated
statistical figures of Eurostat. Projection of on-site CHP, distinctly from CHP with
steam distribution, is produced by the PRIMES power/steam model, using
assumptions regarding the evolution of steam selling market as part of the overall
industrial steam consumption.

Calculations For this purpose, the fuels, which are inputs to CHP plants, split between a part
corresponding to an equivalent electricity-only plant and another part corresponding
consistent to steam output by the CHP plant. The calculation uses a formula, which is similar to
with CHP that used by the guidelines implementing the CHP Directive using distinct parameters
by type of CHP technology. The fuel to steam ratios as produced by this calculation are
Directive
usually higher than one, because steam is a by-product of efficient CHP technologies
and the additional fuel amount needed to produce steam on top of fuels used for the
equivalent electricity-only plant is much lower than fuel amounts used by boilers to
produce the same amount of steam.
This calculation method allows PRIMES to compute in projections the ratio indicating
the share of electricity and steam produced by CHP by including only the part of CHP,
which complies with the efficiency criteria for CHP, which are prescribed in the CHP
Directive. Thereby, the model can also evaluate energy and cost impacts of imposing
policy-driven targets concerning the share of efficient CHP in the future.
 Energy Forms in PRIMES Energy Balances
Solids
hard coal
patent fuels
coke
tar, pitch, benzol
lignite
other solids
Crude oil
Feedstock to refineries
Liquids
refinery gas
liquefied petroleum gas
gasoline
kerosene
naphtha
diesel oil
fuel oil
other liquids
Gas
natural gas
coke-oven gas
blast furnace gas
gasworks gas
Biomass-waste
Ethanol
Bio-gasoline
Bio-diesel
Bio-kerosene
Bio-heavy
Bio-gas
Solid biomass
Solid waste
Gas waste
Liquid biomass
Nuclear
Hydro
Wind
Solar
Tidal and other renewables
Geothermal heat
Methanol
Hydrogen
Steam/Heat distributed
Electricity
 List of Tables included in the PRIMES Energy Balances by country and by year
Primary energy
Primary production
Recovery from coal liquefaction plants
Recovery from gas to liquids plants
Recovery of gas from blending of various methane
Net imports
Stock changes (+ or -)
Bunkers
Gross inland consumption
Total transformation input
Transformation input in thermal power stations
Transformation input in nuclear power stations
Transformation input in district heating plants
Transformation input for production of hydrogen and biofuels
Transformation input in patent fuel and briquetting plants
Transformation input in coke-oven plants
Transformation input in blast furnace plants
Transformation input in gas works
Transformation input in refineries
Transformation input in hydrogen production
Transformation input in methanol production
Transformation input in ethanol production
Transformation input in charcoal production
Transformation input in for blended natural gas
Transformation input in coal liquefaction plants
Transformation input in gas-to-liquids (GTL) plants
Total transformation output
Transformation output from thermal power stations
Transformation output of nuclear power stations
Transformation output from district heating plants
Transformation output of hydrogen and biofuels
Transformation output of patent fuel and briquetting plants
Transformation output from coke oven plants
Transformation output from blast furnace plants
Transformation output from gasworks
Transformation output from refineries
Transformation output from hydrogen production
Transformation output from methanol production
Transformation output from ethanol production
Transformation output from charcoal production
Inter-product exchanges and transfers
 List of Tables included in the PRIMES Energy Balances by country and by year
Total consumption of the energy branch
Energy branch consumption - own consumption and pumping in power generation
Energy branch consumption - refineries
Energy branch consumption - other sectors
Consumption in Charcoal production plants (Energy)
Consumption in Gas-to-liquids (GTL) plants (Energy)
Consumption in Gasification plants for biogas
Consumption in gas system (storage, LNG, etc. except pipeline gas)
Consumption in Coal Liquefaction Plants
Consumption in Oil and gas extraction
Energy Sector Consumption Coke-Oven & Gas-Works Plants
Energy Sector Consumption Mines & Patent Fuel/Briquetting plants
Consumption in Nuclear industry
Pumped storage power stations balance (derived aggregate)
Own Use in Electricity, CHP and Heat Plants
Distribution losses
Available for final consumption
Non-energy consumption
Final non energy consumption
Final non energy consumption in petrochemicals
Final non energy consumption in other sectors
Final energy consumption in industry
Final energy consumption in iron and steel
Final energy consumption in non-ferrous metal industry
Final energy consumption in chemical industry
Final energy consumption in glass, pottery and building materials industry
Final energy consumption in paper and printing industry
Final energy consumption in food, drink & tobacco industry
Final energy consumption in textile, leather & clothing industry
Final energy consumption in engineering and other metal industry
Final energy consumption in other industries
Final energy consumption in transport
Final energy consumption in railways
Final energy consumption in road transport
Final energy consumption in air transport
Final energy consumption in inland navigation
Final energy consumption in Pipeline transport
Final energy consumption in other transport (port cranes, airport vehicles, etc.)
Final energy consumption in households, services. etc.
Final energy consumption in households
Final energy consumption in services
Final energy consumption in agriculture
Statistical differences
 Q. The PRIMES Biomass model
Q.1. Model scope and aim
The PRIMES Biomass model is a modelling tool aimed at contributing to the energy
system projections for the EU Member-States and the impact assessment of policies
promoting renewable energy sources and addressing climate change mitigation. The
detailed numerical model simulates the economics of supply of biomass and waste for
energy purposes through a network of processes, current and future, which are
represented at a certain level of engineering detail for which a very detailed database
of biomass and waste processing technologies and primary resources has been
developed.
The model transforms biomass feedstock –therefore primary energy- into bio-energy
commodities –secondary or final form- which undergo further transformation in the
energy system, e.g. as input into power plants, heating boilers or as fuels for
transportation.
The model calculates the inputs in terms of primary feedstock of biomass and waste
to satisfy a given demand for bio-energy commodities; the model further estimates
the land use and the imports necessary and provides quantification of the amount of
production capacity required. Furthermore, all the costs resulting from the
production of bio-energy commodities and the resulting prices of the commodities are
quantified.
The model covers all EU27 Member States individually and covers the entire period
from 2000 to 2050 in five-year periods. It is calibrated to Eurostat statistics wherever
possible for the years 2000 to 2010. Data from Eurostat is complemented by other
statistical sources to fill in the database necessary for the model to function.
The model can operate as a standalone model if the demand for bio-energy
commodities is given exogenously, but is more often used together with the PRIMES
Energy System Model as a closed loop system.
The PRIMES Biomass model is developed and maintained at E3modelling. The model
databases were improved over the years and were recently harmonised with other
European models within the Biomass Futures project. The current model version has
been thoroughly updated in autumn 2011 where the technology and process database
was updated. The historical data are updated to the latest available 2010 statistics.

Q.2. Structure, feedstock and conversion technologies

Q.2.a. Structure
The general structure of the model can be described in the following way:
PRIMES has
devoted much Step 1: A primary biomass commodity (e.g., sugar, starch etc) is produced/derived
focus on biomass from the primary resource (e.g. energy crops) through a primary transformation stage
supply because
technology (e.g. cultivation).
transformation in
this sector and Step 2: The primary commodity is then, passed through a pre-processing stage (e.g.
supply potential are drying) that produces a secondary/intermediate commodity.
important for
sustainability goals.
Step 3: The secondary commodity is the input to the transformation process from
which the final energy product (e.g. biofuel) is derived. Logistics are taken into
account as part of the different processes. Final bio-energy supply exactly matches
demand derived from the rest of PRIMES models.

Q.2.b. Feedstock
The primary production of biomass has been classified into the following categories:
energy crops, agricultural, forestry and industrial waste and aquatic biomass (i.e.
algae). Depending on the type of the plants that are cultivated, energy crops are
further distinguished into starch, sugar, oil and lignocellulosic crops. This
classification is dictated by the differentiation of the methods that each plant category
may be processed with and the final products that derive from them. Starch crops
include resources such as maize, wheat, barley etc., sugar crops refer mainly to sugar
beet and sweet sorghum and oil crops consist of rapeseed, sunflower seed, olive
kernel etc. Regarding lignocellulosic crops there is a distinction between wood crops,
such as poplar, willow etc., and herbaceous lignocellulosic crops like miscanthus,
switch grass, reed etc.
Forestry is split into wood platform, i.e. organised and controlled cutting of whole
trees for energy use, and wood residues, i.e. the collecting of forestry residues only.
As mentioned above several types of wastes have been identified as potential sources
for energy supply. These include industrial solid waste, pulp industry waste (black
liquor), used oils and fats, municipal waste, sewage sludge, landfill gas, manure and
animal wastes. The table below summarises all the types of primary biomass/waste
effectively used in the model.

Energy Crops Forestry Wastes and Aquatic Biomass

Residues
Starch Crops Wood Platform Agricultural Residues Algae Biomass
Sugar Crops Forest Residues Wood Waste
Oil Crops Waste Industrial Solid
Lignocellulosic Crops Black Liquor
Used oils and fats
 Herbaceous Crops
Municipal Waste
 Wood Crops
Sewage Sludge
Landfill Gas
Manure
Animal Waste
Q.2.c. Biomass Conversion
The PRIMES Biomass model includes numerous production pathways for the
production of biofuels for transportation, both for road and non-road, as well as
pathways producing bio-energy commodities as inputs into electricity and heat
generation sectors. The end products available in the model include bio-energy
commodities such as biofuels for road transportation, biogas, small scale solids
(mainly pellets) and large scale solids, which mainly are for use in the power
generation and industry. Transportation biofuels include diesel and gasoline from
biomass (both conventional and advanced biofuels), bio-kerosene for aviation, bio-
heavy for navigation, as well as bio-gas. For gasoline and diesel the model
differentiates between conventional and advanced biofuels, which are considered to
be fully fungible with conventional fuels and can therefore be used in existing engines.
For gaseous bio-energy commodities, the model differentiates between bio-methane,
which is biogas upgraded to pipeline quality and biogas.
Some of the bio-energy production technologies, such as fermentation of sugars for
ethanol production or transesterification of vegetable oil for the production of
biodiesel, are technologically and economically mature processes and are already well
established in Europe for the production of biofuels. Other technologies, such as
pyrolysis of wood, offer significant benefits, regarding mainly the utilisation of
cheaper and abundant feedstock, require further research and development to
become economically competitive. This process is fully endogenous in the PRIMES
biomass model.
An extensive literature review was conducted in order to identify those biomass-to-
bio-energy commodities conversion technologies that bear significant potential for
future penetration in the biofuel market. The choice of the technologies that are finally
selected to be included in the PRIMES biomass model was made based on the current
status of technical and economic development, research efforts and possibilities for
future improvements, type of feedstock and type and characteristics of final products.
The technologies that are incorporated in the model are based on the different
conversion chains which constitute pathways of primary biomass transformation to
ready-to-use bio-energy commodities.
FEEDSTOCK PRODUCTION PATHWAY END PRODUCT
Starch crops, Sugar crops Fermentation Ethanol
Woody Biomass Enzymatic Hydrolysis and Fermentation Cellulosic Ethanol
Enzymatic Hydrolysis and Fermentation
HTU process, deoxygenation and upgrading
Woody Biomass
Pyrolysis, deoxygenation and upgrading Ethanol
Pyrolysis, Gasification, FT and upgrading (advanced)
Woody Biomass, Black Liquor Gasification, FT and upgrading
Aquatic Biomass Transesterification, Hydrogenation and Upgrading
Oil crops Transesterification
Biodiesel
Starch crops, Sugar crops Enzymatic Hydrolysis and deoxygenation
Oil crops Hydrotreatment and deoxygenation
Woody biomass, Black Liquor Gasification and FT
Aquatic Biomass Transesterification and Hydrogenation Biodiesel
HTU process and deoxygenation (advanced)
Woody biomass Pyrolysis and deoxygenation
Pyrolysis, Gasification and FT
Gasification and FT
HTU process and deoxygenation
Woody biomass
Pyrolysis and deoxygenation Bio-kerosene
Pyrolysis, Gasification and FT
Aquatic Biomass Transesterification and Hydrogenation
Woody biomass Gasification and methanol Synthesis Bio-methanol
Woody biomass Gasification and DME Synthesis Bio-DME
Woody biomass, Black Liquor Gasification
Organic Wastes, Starch Anaerobic Digestion
Biogas/ Bio-methane
Enzymatic Hydrolysis
Woody biomass
Catalytic Hydrothermal Gasification
Hydrothermal Upgrading (HTU process)
Woody biomass
Pyrolysis Bio Heavy Fuel Oil
Black Liquor Catalytic Upgrading of black liquor
Landfill, Sewage Sludge Landfill and sewage sludge
Waste Gas
Organic Wastes Anaerobic Digestion
Industrial Waste, Municipal Waste
(solid) RDF Waste Solid
Woody biomass Small Scale Solid
Woody biomass Large Scale Solid
Q.2.d. Technologies for Bioethanol production
Sugars & Starch Bioethanol is used in spark ignition vehicle engines either blended with gasoline or in
pure form if the engines are properly modified. At present bioethanol is mainly
Fermentation produced from sugar crops via fermentation. Currently in Europe sugar beet and
sweet sorghum are mainly used as feedstock. Starch crops are also being used as
feedstock. In that case an additional pre-process stage is needed to hydrolyse starch
into simpler sugars before fermentation that implies a cost difference between sugar
& starch fermentation processes. Thus, in PRIMES Biomass Model Sugar Fermentation
& Starch Fermentation are treated as separate technologies.

Lignocellulosic Bioethanol can also be produced using as feedstock lignocellulosic crops through the
biochemical conversion of the cellulose and hemicelluloses components of biomass
Fermentation feedstock into fermentable sugars. Cellulosic ethanol has the potential to perform
better in terms of energy balance, greenhouse gas (GHG) emissions and land-use
requirements than starch-based biofuels. Unlike production of bio-ethanol from sugar
 and starch crops, this process is still under development however, a lot of research is
taking place both in Europe and in the USA implying significant potential.

Q.2.e. Technologies for Biodiesel production

Transesterification Biodiesel is merely produced from vegetable oils by catalytic transesterification with
methanol. Biodiesel produced in this way has similar properties with fossil diesel and
may be used in conventional engines blended up to a proportion with fossil diesel or
in modified engines in higher proportions. Vegetable oils may be produced from
several biomass sources, such as rapeseed, soya been, sunflower, olive kernel etc. In
Europe the most common feedstock for the production of vegetable oil as feedstock
for further conversion into biodiesel is rapeseed. Other vegetable and animal fats as
well as used oils may also be used as feedstock to the transesterification process.
Transesterification is a well-established technology and is largely deployed in Europe.
Additionally, research has been performed to examine algae oil production from
microalgae cultivation that could be used as feedstock for the production of biodiesel
via transesterification offering various potential advantages when compared with
traditional oil crops.

Fischer Tropsch The production of diesel from coal via the Fischer-Tropsch process is a technology
with long history. Historically the approach has focused on the conversion of coal-to-
Synthesis liquid fuels and chemicals. Recently the utilisation of biomass derived syngas is
proposed for the production of Fischer-Tropsch biodiesel. The thermo-chemical route
involves the production of a synthesis gas, which is cleaned, before passed through
the Fischer-Tropsch process, to create a range of liquid fuels, but primarily synthetic
diesel. The production of Fischer-Tropsch diesel requires thorough cleaning and
conditioning of the biomass derived syngas, which currently bears a lot of technical
difficulties and challenges and deteriorates the economics of the technology. However,
the combination of the multiple feedstock gasification with the synthesis of Fischer-
Tropsch diesel is an attractive alternative for the production of a fossil diesel
substitute.

Pyrolysis Pyrolysis oil is produced by a thermo-chemical conversion process called flash

pyrolysis. In order to be used as transport fuel, pyrolysis oil has to be hydro-
deoxygenated using catalysts and stabilised to reach specific quality requirements.
Since it is not mixable with fossil diesel, the resulting fuel may only be used directly in
modified diesel engines. Pyrolysis oil can also be used for co-firing in power and
steam generating units, or may be gasified for the production of syngas. The
technology has not reached to a maturity status yet and there are significant
difficulties that have to be overcome. However, since almost any type of biomass can
be used in flash pyrolysis, including lignocellulosic biomass, the technology is
attractive and bears significant potential for future deployment.

Hydro Thermal Another substitute of fossil diesel is proposed by converting almost all types of
biomass into liquid biofuel via a process called hydro-thermal upgrading (HTU).
Upgrade During HTU process, the biomass is decomposed in water to produce a crude oil-like
liquid called ‘bio-crude’. The resulting ‘bio-crude’ is further upgraded through
hydrogenation with catalysts to achieve fossil diesel quality and may be blended in
any proportion with conventional fossil diesel. The technological status of the HTU
process has not reached maturity yet. Furthermore, it is a highly energy intensive
process which further reduces its economic performance. Nevertheless, the utilization
of a wide variety of feedstock ranks HTU as a candidate technology for future
production of biodiesel.

Q.2.f. Technologies for bio-kerosene production

Bio-kerosene is alternative for jet fuel having similar properties to petroleum-derived
kerosene. Currently bio-kerosene production is under research and only test flights
have been performed. The airline industry aims not only at replacing fossil with
 renewable fuels but also to improve fuel efficiency standards and reduce the volume
of greenhouse gas emissions. PRIMES includes fungible bio-kerosene production in
the period mainly after 2030.

Q.2.g. Technologies for biogas and bio-methane production

Anaerobic A series of biological processes in which microorganisms break down biodegradable
material (biomass); in the absence of oxygen, anaerobic bacteria ferment biomass into
Digestion biogas. Biogas can be produced this way from almost any organic matter such as
agricultural residues, animal waste and manure. In the PRIMES Biomass Model,
Anaerobic Digestion is used to produce biogas from every raw material mentioned
above. Biogas is a mixture of CO2 and CH4. Methane represents approximately a 60%
in the total mixture. In order to increase CH4 proportion in the mixture, biogas passes
through an upgrading process where CO2 is absorbed or scrubbed and finally leaves
98% of bio-methane that can be directly injected into the natural gas grid.

Gasification Biogas can also be produced via gasification. This route has a big potential as a wider
range of feedstock like wood can be used. Biomass is gasified at high temperature,
producing bio-syngas. The bio-syngas enters a gas cleaning section and then passes
through a methanation unit where CO and H2 are converted into bio-methane and
CO2. After CO2 removal, the gas is ready for injection into the natural gas grid.

Waste Gas Waste gas consists of Sewage Sludge Gas and Landfill Gas that can be produced via
anaerobic digestion technology using Waste Sewage Sludge and Waste Landfill Gas as
production feedstock. Anaerobic digestion for the production of waste gas is currently widely
used in Europe. Due to the impurity and the lower methane content of waste gas
compared to that of the biogas (bio-methane) described above, waste gas cannot be
injected into the natural gas grid and its main applications are to produce electricity
and heating at small scale.

Q.2.h. Technologies for bio-heavy production

Bio-heavy includes ’bio-crude’ and ’pyrolysis oil’ produced via Pyrolysis and
Hydrothermal Upgrade. It is mainly produced to be further altered into biodiesel
through transesterification, but could also be used for heat generation or as transport
fuel in bunkers.

Q.2.i. Technologies for Small & Large Scale production of solids

Small & Large Scale Solid (woody) consists of wood logs and pellets for small and
Solid woody large-scale combustion for power and heating generation, produced from pelletizing
biomass and logging processes of wood biomass.

Waste Solid Waste solid consists of Mass burn waste (MBW) and Refused derived fuel (RDF). Mass
burn refers to the incineration of unsorted municipal waste in a Municipal Waste
production Combustor (MWC) or other incinerators designated to burn only waste from
municipalities. RDF is a solid fuel for direct combustion and covers a wide range of
waste materials processed to fulfil guideline, regulatory or industry specifications
mainly to achieve a high calorific value. Waste derived fuels include residues from
MBSW recycling, industrial waste, industrial hazardous waste, biomass waste, etc.
RDF can be produced from municipal solid waste through a number of different
processes that in general consist of separation and sorting, size reduction (by
shredding, chipping and milling), drying and finally transforming the combustible
waste into cylindrical solid fuel.

Q.2.j. Technologies for bio-hydrogen production

Bio-hydrogen is a promising future energy source due to its very high-energy content
and the fact that it produces almost no emissions when burnt. It could perform either
as direct fuel in engines that would burn pure hydrogen or as electric power source
for electric motor vehicles (through a fuel cell). Bio-hydrogen can be produced from
 bio-syngas, a mixture of H2 and CO formed from biomass derived char, oil or gas. In
PRIMES Biomass model, bio-syngas is derived through biomass gasification, to
achieve higher ratios of H2/CO, an important factor that affects its performance as fuel
source. The resulting mixture passes then through a solvent separation system to
absorb CO and release bio-hydrogen.

Q.3. Biomass Model Methodology

The PRIMES Biomass model is an economic supply model that computes the optimal
use of resources and investment in biomass transformation processes, to meet a given
demand for final biomass energy products under least cost conditions. The PRIMES
Biomass model solves a non-linear optimisation model. Concatenated with the rest of
the PRIMES suite, establishing a closed loop, the PRIMES Biomass model uses a Mixed
Complementarity Problem (MCP) algorithm to determine the equilibrium of demand
and supply, and the prices of bio-energy commodities. The time horizon of the model
is 2050. The model provides dynamic projections to the future from 2015 until 2050
in 5-year time periods with years 2000 to 2010 being calibration years.

The PRIMES biomass model solves for cost minimisation from the perspective of a biomass supply planner,
who fully anticipates demand, fuel prices, biomass costs and technology improvement potentials if deployed in
large scale. The optimisation is constrained by:
a) the graph of possible conversions and transformations of feedstock to final bio-energy commodities,
b) demand for bio-energy commodities,
c) availability of land and feedstock,
d) cost-supply curves denoting import possibilities and
e) policy regulations including sustainability and fuel quality criteria.
The model determines:
a) the optimal use of biomass/waste resources,
b) the investments in technologies for biomass conversion to bio-energy commodities,
c) the use of land,
d) the imports from outside the EU and the intra-EU trade of feedstock and bio-energy commodities,
e) the costs and the consumer prices of the final bio-energy products as well as
f) the greenhouse gas (GHG) emissions resulting from the bio-energy commodities life-cycle.
The decision on investment for the secondary and final transformation processes is endogenous using
technology vintages and dynamics of technology development. Furthermore, endogenous learning-by-doing
for all technologies has been included, to simulate technological change and decrease of costs of technologies
as related to the cumulative experience gained in the process of commodities production. Improvements in
each technology are described by one learning-by-doing curve for each technology, uniform for all Member
States of the EU; therefore learning-by-doing effects spill over to the whole EU.

Q.3.a. Inputs
Overview of The model input data are country specific data. They include data on:

input data  agricultural land use productivities and availability,

 costs and commodity prices (prices of electricity, gas and other liquid fuels
come from the rest of the PRIMES model),

 technical-economic features of biomass conversion processes including

learning by doing potential

 biomass and waste resources potential and the cost supply curves,

 import and bilateral trade possibilities.

 For the feedstock prices, the model uses cost-supply curves (with increasing slopes)
which are specific by country and depend on land availability, productivity trends and
the use of fertilizers (counted in the external effects). Exogenous assumptions and
estimates are used about land availability and yield improvement possibilities for the
various energy crops. The yields are assumed to increase over time due to technology
developments and depending on deployment of specific agricultural policies, which
vary by scenario.

Q.3.b. Endogenous trade

Endogenous The model fully formulates trade of both primary biomass and end bio-energy
commodities within the EU and with regions outside the EU. Tradable feedstock
trade of considered are pure vegetable oil, which is mainly imported palm oil, and solid
feedstock and biomass. The bio-energy products assumed tradable are solid biomass, conventional
and next generation biodiesel, bioethanol, bio-gasoline (meaning cellulosic
bio-energy
bioethanol) and bio-kerosene. The trade takes place both between EU Member States
commodities with endogenous transportation costs depending on transportation possibilities and
with other countries outside the EU. The regions outside the EU carrying biomass
trading with the EU are modelled in aggregate categories including North America, CIS
and the rest of the world. The trade that takes place between Europe and the rest of
the world includes as main providers for wood CIS and North America, while for
sugarcane bio-ethanol Brazil. Imported oil is for the most part palm oil mainly from
Indonesia and Malaysia. Import possibilities from outside the EU are described
through cost-supply curves, which change over time and in the context of different
scenarios depending on assumptions about biomass use for energy purposes in the
rest of the world. Trade within the EU depends on transportation costs, which are
determined, based on aggregate spatial information.

Inputs

Historical Data
PRIMES Biomass
Model

Country specific Data

Fulfil restrictions

PRIMES
Model Technoeconomic Data

Outputs:
Energy production
Cost function flows, Import flows,
minimization capacities, commodity
costs etc.
Imports/Exports Data

Q.3.c. Mathematical specification of the model

The biomass supply model of PRIMES solves a problem of minimizing total long-term
supply costs of meeting a given demand for bio-energy commodities, which is derived
from the rest of PRIMES model. The minimization is subject to equilibrium
constraints, which represent the cost structure of various feedstock supplying
possibilities, as well as the cost functions of technology suppliers. Policy-related
restrictions are represented as overall constraints, such as for example the
sustainability criteria. It is assumed that a variety of biomass producers and
transformers acting in all EU Member-States compete with each other in production
and in biomass commodity trading among the Member-States. Thus, the optimization
solves for all the Member-States simultaneously as well as for the entire time horizon
assuming perfect anticipation by all market actors. The model also determines bio-
energy commodity prices because of maximizing social surplus by Member-State
subject to recovering all types of fixed and variable costs of biomass supply.
At a first glance the biomass supply optimization resembles a least cost transport
problem consisting of finding the least cost way of meeting demand for bio-energy
commodities (denoted by 𝑖) through feedstock resources (denoted by 𝑠) which are
stepwise transformed into final commodities in a variety of processes (denoted by 𝑗).
The technically feasible conversion and transformation pathways are considered to
belong to the mapping ℎ(𝑠, 𝑗, 𝑖). Both demand and supply are located at the EU
Member-States (denoted by 𝑛), which are linked together through a transportation
network used for trading bio-energy commodities among the Member-States. In
addition, the Member-States are connected to non-EU regions for importing biomass
feedstock and/or ready-made bio-energy commodities.
Feedstock can be produced in the EU from crops, residues (agriculture, forestry), and
wastes. Owners of resources used to produce feedstock (such as land, residue, or
waste collectors) are assumed to have different cost structures and to compete with
each other. Thus, supply of feedstock is assumed to derive from cost-supply curves,
denoted by 𝑓𝑠 (𝐹𝑠 ), which depend on quantities produced annually (𝐹𝑠 ) and exhibit
decreasing returns to scale. The cost supply curves are also specified by Member-State
(𝑛) and over time (𝑡).
Exporters from outside the EU addressing the Member-State markets are assumed to
price feedstock or bio-energy commodities according to their own cost-supply
structure, which is subject to resource limitations. Thus, import prices increase with
imported quantities (𝑠 𝑜𝑟 𝑖) following an ascending cost-supply function:
𝑚𝑘,(𝑠 𝑜𝑟 𝑖) (𝑀𝑘,(𝑠 𝑜𝑟 𝑖) ) where 𝑀 denotes imported quantities and 𝑘 denotes the various
non-EU importing origins.
The processes (𝑗) transforming feedstock into bio-energy commodities have
processing capacities (𝐾𝑗,𝑡 ) which are formed by accumulating investment (𝐼𝑗,𝑡 ). The
technology characteristics (unit costs, efficiency, and input/output ratios) are specific
to the year of investment, but for new installations they evolve over time depending
on technology supply which follows learning-by-doing curves, denoted
by ℓ𝑗 (∑𝑛 ∑𝑡 𝐼𝑗,𝑛,𝑡 ) exhibiting decreasing costs and increasing performance as a
function of total installed capacity in the EU.
Production in time 𝑡 from a processing unit 𝑗 built in time 𝜏 (i.e. 𝐺𝑗,𝜏,𝑡 ) is constrained
by available capacity (𝐾𝑗,𝜏 ). In addition, the rate of use of capacities cannot decrease
below a certain level, otherwise, the capacity is not at all used. As the optimization
assumes perfect foresight, obviously only capacities with sufficiently high rates of use
will be built. Thus, the model simulates competition between various processing
technologies.
Both the decreasing costs due to the learning curves and the capacity usage
constraints violate standard convexity requirements and so the optimization is
formulated as a mixed-integer programming problem.
The main unknown variables are: 𝐹𝑠,𝑛,𝑡 the domestic production of feedstock, 𝐺𝑖,𝑗,𝑛,𝜏,𝑡
the production of processes, 𝑀𝑘,𝑠,𝑛,𝑡 and 𝑀𝑘,𝑖,𝑛,𝑡 the imports of feedstock and ready-
made bio-energy commodities from non EU countries, 𝐾𝑗,𝑛,𝜏 and 𝐼𝑗,𝑛,𝜏 the capacity and
investment in processes (of vintage 𝜏) and 𝑋𝑖,𝑛,𝑛𝑛,𝑡 the exchanges of bio-energy
commodities between the EU Member-States (𝑛𝑛 being an alias of 𝑛).
Market equilibrium is formulated by Member-State and for each bio-energy
commodity, requiring that total supply from domestic production and imports (both
outside the EU and from other EU countries) meets exactly given demand in each
period. Market equilibrium is thus ensured through the following condition:

∑ ∑ 𝑮𝒊,𝒋,𝒏,𝝉,𝒕 + ∑ 𝑴𝒌,𝒊,𝒏,𝒕 + ∑(𝑿𝒊,𝒏𝒏,𝒏,𝒕 − 𝑿𝒊,𝒏,𝒏𝒏,𝒕 ) = 𝒅𝒊,𝒏,𝒕 ∀𝒏, ∀𝒕 (1)

𝒋 𝝉≤𝒕 𝒌 𝒏𝒏

where 𝑑𝑖,𝑡 is the demand for bio-energy commodities, given from the core PRIMES
model.
Production by transformation processes uses inputs and outputs related to each other
through a production possibility function, denoted by ℊ𝑠,𝑗,𝑖 which determines demand
(𝑓) (𝑒)
for feedstock (𝐺𝑠,𝑗,𝑖,𝑛,𝜏,𝑡 ) and fuel (and electricity) consumption (𝐺𝑠,𝑗,𝑖,𝑛,𝜏,𝑡 ):

(𝒇)
𝑮𝒔,𝒋,𝒊,𝒏,𝝉,𝒕 = 𝓰𝒔,𝒋,𝒊 (𝑮𝒊,𝒋,𝒏,𝝉,𝒕 ) ( ∀𝒔, ∀𝒋, ∀𝒊) ∈ 𝒉(𝒔, 𝒋, 𝒊) 𝒂𝒏𝒅 ∀𝒏, ∀𝒕, ∀𝝉 ≤ 𝒕 (2)

(𝑒)
𝐺𝑠,𝑗,𝑖,𝑛,𝜏,𝑡 = ℊ𝑠,𝑗,𝑖 (𝐺𝑖,𝑗,𝑛,𝜏,𝑡 ) ( ∀𝑠, ∀𝑗, ∀𝑖) ∈ ℎ(𝑠, 𝑗, 𝑖) 𝑎𝑛𝑑 ∀𝑛, ∀𝑡, ∀𝜏 ≤ 𝑡 (3)

Capacities of processing are determined by investment accumulation, as follows

(initial conditions concerning old existing capacities are not shown):

𝑮𝒊,𝒋,𝒏,𝝉,𝒕 ≤ 𝑲𝒋,𝒏,𝝉,𝒕 ∀𝒊, ∀𝒋, ∀𝒏, ∀𝒕, ∀𝝉 ≤ 𝒕 (4)

𝐺𝑖,𝑗,𝑛,𝜏,𝑡 ≥ 𝑢𝑗,𝑛,𝜏 𝐾𝑗,𝑛,𝜏,𝑡 𝑜𝑟 𝐺𝑖,𝑗,𝑛,𝜏,𝑡 = 0 ∀𝑖, ∀𝑗, ∀𝑛, ∀𝑡, ∀𝜏 ≤ 𝑡 (5)

𝐾𝑗,𝑛,𝜏,𝑡 = 𝐼𝑗,𝑛,𝜏 − 𝐷𝑗,𝑛,𝜏,𝑡 ∀𝑗, ∀𝑛, ∀𝑡, ∀𝜏 ≤ 𝑡 (6)

where 𝑢𝑗,𝑛,𝜏 is the minimum rate of use of capacities and 𝐷𝑗,𝑛,𝜏,𝑡 denotes the
decommissioned capacities, which depend on technical lifetime. The time index 𝜏,
which must be lower or equal than current projection time 𝑡, denotes the technology
vintage for processing units and so production as well as technology characteristics
are specific to a vintage.
Total demand for feedstock by type has to be met by domestic production and by
imports from non-EU countries:

(𝒇)
∑ 𝑴𝒌,𝒔,𝒏,𝒕 + 𝑭𝒔,𝒏,𝒕 = ∑ ∑ 𝑮𝒔,𝒋,𝒊,𝒏,𝝉,𝒕 ∀𝒔, ∀𝒏, ∀𝒕 (7)
𝒌 (𝒋,𝒊)∈𝒉(𝒔,𝒋,𝒊) ∀𝝉≤𝒕

The part of domestic feedstock originating from crops is associated with land use
(𝐿𝑠,𝑛,𝑡 ) through a production function (𝑦𝑠,𝑛 ) which exogenously assumes yield growth
trends, specifically by crop type and by country. Similarly, other feedstock types, such
as residues or waste, are using primary resources, which are also denoted by, 𝐿𝑠,𝑛,𝑡
and their use depends on productivity trends captured through the function 𝑦𝑠,𝑛 . Total
domestic resources by type of feedstock give upper bounds𝐿̅𝑠,𝑛,𝑡 , which represent
technical potentials.

𝑳𝒔,𝒏,𝒕 = 𝒚𝒔,𝒏 (𝑭𝒔,𝒏,𝒕 ) ∀𝒔, ∀𝒏, ∀𝒕 (8)

𝐿𝑠,𝑛,𝑡 ≤ 𝐿̅𝑠,𝑛,𝑡 ∀𝑠, ∀𝑛, ∀𝑡 (9)

Emissions of greenhouse gases are related to energy consumption in the
transformation processes, which include collection and transportation of feedstock,
and to emissions related to domestic production of crops. Emissions by type of bio-
energy commodities across the chain of production will have to be lower than a
threshold (a sustainability criterion):

(𝒆) (𝒇)
∑ ∑ 𝒆𝒎(𝒆) ∙ 𝑮𝒔,𝒋,𝒊,𝒏,𝝉,𝒕 + ∑ 𝒆𝒎(𝒇) ∙ 𝒂𝒔,𝒋,𝒊 ∙ ∑ 𝑮𝒔,𝒋,𝒊,𝒏,𝝉,𝒕
(𝒔,𝒋)∈𝒉(𝒔,𝒋,𝒊) ∀𝝉≤𝒕 (𝒔,𝒋)∈𝒉(𝒔,𝒋,𝒊) ∀𝝉≤𝒕 (10)
≤ 𝒙𝒆𝒊,𝒏,𝒕 ∙ 𝒅𝒊,𝒏,𝒕 ∀𝒊, ∀𝒏, ∀𝒕
where 𝑒𝑚 are greenhouse gas emission factors, 𝑥𝑒 is the specific emission threshold
(the sustainability criterion) and 𝑎𝑠,𝑗,𝑖 denote the share of feedstock of type 𝑠 used to
produce bio-energy commodity 𝑖 through a process of type 𝑗.

It is assumed that the actors optimizing total biomass supply anticipate the economics
of feedstock supply, as well as the cost functions of imports and the learning curves of
technology supply. Thus, they take into account the gradients of the corresponding
cost-supply curves in their optimization. In this sense, the optimization corresponds
to a problem of mathematical programming with equilibrium constraints.
Total biomass supply system cost include in addition the annuity payments for capital
investment in transformation processes, the variable and energy costs, the fixed
operation and maintenance costs and the transportation costs which depend on
distances between the Member-States. The annuity payments depend on a weighted
average cost of capital (𝜌𝑗,𝑛 ) which may differ by country and by type of process. The
aggregation of total costs over time use present values discounted using a social
discount rate (𝛿).
Total intertemporal biomass supply system cost is then defined as follows:

𝑪𝒐𝒔𝒕 = ∑ 𝒆−𝜹∙𝒕 ∑ (∑ (∑ 𝒇𝒔 (𝑭𝒔,𝒏,𝒕 ) + ∑ 𝒎𝒌,𝒔 (𝑴𝒌,𝒔,𝒏,𝒕 ))

𝒕=𝟏 𝒏 𝒔 𝒇 𝒌

+ ∑ ∑ 𝒏𝒋,𝒏 (𝝆𝒋,𝒏 ) ∙ 𝓵𝒋 (∑ ∑ 𝑰𝒋,𝒏𝒏,𝒕𝒕 ) ∙ 𝑰𝒋,𝒏,𝝉

𝒋 𝝉≤𝒕 𝒏𝒏 𝒕𝒕 (11)
+ ∑ ∑ ∑ 𝒗𝒊,𝒋,𝒏 (𝑮𝒊,𝒋,𝒏,𝝉,𝒕 )
𝒋 𝒊 𝝉≤𝒕

+ ∑ (∑ 𝒓𝒊,𝒏,𝒕 (𝑿𝒊,𝒏𝒏,𝒏,𝒕 ) + ∑ 𝒎𝒌,𝒊 (𝑴𝒌,𝒊,𝒏,𝒕 )))

𝒊 𝒏𝒏 𝒌

where 𝑓𝑠 , 𝑚𝑘,𝑠 , ℓj , 𝑣𝑖,𝑗,𝑛 , 𝑟𝑖,𝑛,𝑡 , 𝑚𝑘,𝑖 denote respectively the feedstock cost-supply
function, the imported feedstock cost-supply function, the technology learning-by-
doing function, the variable cost function for processes, the transportation cost
function for intra-EU trade and the cost-supply function for imported bio-energy
commodities from outside the EU. Annuity payment factors for capital investment are
represented by 𝑛𝑗,𝑛 (𝜌𝑗,𝑛 ).

The optimization problem consists in minimizing total cost given by (11), subject to
the constraints that are described by (1) to (10) and to non-negativity constraints for
the unknown variables. The anticipation of the equilibrium conditions by the cost-
minimizing agent is incorporated directly in the objective function through the cost-
supply curves and so it is not needed to solve the model using an MPEC16 algorithm.

16 Mathematical programming with equilibrium constraints (MPEC) is the study of constrained

optimization problems where the constraints include variational inequalities or complementarities. The
lower level optimization problems represent decision by suppliers (of feedstock, imports and technology)
for which it is assumed that they have analytical solutions in the form of cost-supply functions.
The dual variable of (1) is the long-term marginal cost of demand for bio-energy
commodities and the dual variable of (7) is the long-term marginal value of feedstock
supply.
To determine the prices of bio-energy commodity by type and by country, the model
formulates a Ramsey-Boiteux pricing methodology. This rule takes the perspective of
a multiproduct monopolist, which sets the prices to maximize social surplus subject to
a constraint on profits for which total costs include fixed and sunk costs. Such a rule
often applies to regulated utilities, which develop new infrastructures, and is
consistent with regulators aim at maximizing welfare together with ensuring effective
investment. For long term planning, as it is the purpose of the model, the Ramsey-
Boiteux pricing is appropriate for an emerging industry, such as biomass production
for energy purposes. The price-setting outcome is also compatible with well-
functioning markets, which will have to be competitive while providing assurance
about fixed cost recovery.
It is assumed that the bio-energy commodities address different markets where they
compete against other forms of energy and that demand for these commodities
depend on prices. The numerical values of the price elasticities are known using the
core PRIMES model, which among others calculates the demand for the bio-energy
commodities. Assume that the implicit demand functions are denoted by 𝑑𝑖,𝑛,𝑡 (𝑝𝑖,𝑛,𝑡 )
where 𝑝𝑖,𝑛,𝑡 are the prices of the bio-energy commodities. Let us denote by
𝜋𝑖,𝑛,𝑡 (𝑑𝑖,𝑛,𝑡 ) the inverse demand functions. Cost of supply of bio-energy commodities
is known by solving the optimization problem mentioned above, which defines an
implicit cost function denoted by 𝐶𝑛,𝑡 (𝑑𝑖,𝑛,𝑡 , ∀𝑖) as depending on the entire bundle of
bio-energy commodity demand by country. Consequently, total revenue 𝑅𝑛,𝑡 by
country, profit Π𝑛,𝑡 and social surplus 𝑊𝑛,𝑡 can be calculated as follows:

𝑹𝒏,𝒕 = ∑ 𝝅𝒊,𝒏,𝒕 (𝒅𝒊,𝒏,𝒕 ) ∙ 𝒅𝒊,𝒏,𝒕 ∀𝒏, ∀𝒕 (12)

𝒊

Π𝑛,𝑡 = 𝑅𝑛,𝑡 − 𝐶𝑛,𝑡 (𝑑𝑖,𝑛,𝑡 , ∀𝑖) − Φ𝑛,𝑡 ∀𝑛, ∀𝑡 (13)

𝑑𝑖,𝑛,𝑡

𝑊𝑛,𝑡 = ∑ ( ∫ 𝜋𝑖,𝑛,𝑡 (𝑧) ∙ 𝑑𝑧) − 𝐶𝑛,𝑡 (𝑑𝑖,𝑛,𝑡 , ∀𝑖) − Φ𝑛,𝑡 ∀𝑛, ∀𝑡 (14)
𝑖 0

Price determination derives from maximizing social surplus 𝑊𝑛,𝑡 calculated by (14),
̅ 𝑛,𝑡 which is typically
subject to profit, calculated by (13), being equal to a fixed value Π
set equal to zero or to an exogenously defined level. Solving this problem leads to
price setting through:

𝒑𝒊,𝒏,𝒕 = 𝝅𝒊,𝒏,𝒕 (𝒅∗𝒊,𝒏,𝒕 ) ∀𝒏, ∀𝒕 (15)

∗
where 𝑑𝑖,𝑛,𝑡 results from social surplus maximization under the profit constraint. The
parameter Φ𝑛,𝑡 may be used to represent a variety of fixed or sunk costs that are
needed to develop the emerging bio-energy market, as well as opportunity costs for
example in relation to prices of energy commodities competing with bio-energy ones.

Q.4. Representation of policies and measures

The PRIMES Biomass model is designed to take into account legislation that concerns
the use of biomass in energy sector and thus constitutes a tool for the evaluation of
the way policies affect the biomass supply system. Additional to current legislation
the model is able to simulate other policy contexts, therefore the impacts of different
policies, beyond current legislation, and measures can be simulated.
 The EU The achievement of the EU 20-20-20 targets is implemented in the PRIMES energy
system model and the demand delivered to the PRIMES Biomass model therefore
Climate & includes these targets. Currently, the model takes into account the RES Directive
Energy (Directive 2009/28/EC), the Fuel Quality Directive (Directive 2009/30/EC) and the
package Biofuels Directive (Directive 2003/30/EC).

Emissions Restrictions per bio-energy commodity express that the greenhouse gas emissions
(GHG) as percentage of emissions avoided for each biomass commodity have to lie
constraints above a certain percentage threshold, determined by current legislation. The
threshold imposed by the RES and the Fuel Quality directives is used, but constraints
that are more stringent can apply depending on the scenario.
Carbon emissions are calculated for each final bio-energy commodity as a sum of
emissions in all stages of the chain of commodity production and transformations. The
emissions are computed by multiplying the quantities of energy forms (oil products,
gas, and electricity) used in the production and transformation of biomass
commodities by specific emissions factors. These factors are obtained from the results
of the rest of PRIMES model.
Emissions resulting from indirect land use change (ILUC) can be included in the
calculation of the overall emissions, despite the fact that ILUC emissions are not taken
into account in current legislation methodologies.

Sustainability Additional to the GHG mitigation criterion, other sustainability related restrictions can
be effectively applied in the PRIMES Biomass model. The criteria currently used in the
criteria
model are the ones set out by the RES and the Fuel Quality directive and are related to
high biodiversity land and to land with high carbon stock. According to legislation, the
raw materials used as biomass feedstock cannot be obtained from high biodiversity
areas, such as undisturbed forests, high biodiversity grasslands and nature protection
areas, unless the production of that raw material is obtained harmlessly. Furthermore,
land with high carbon stock cannot be converted to biomass feedstock cultivation
area, thus areas such as wetlands, continuously forested areas and peat lands are
excluded from the land that can be used for the production of energy crops. These
criteria set restrictions to the total acreage of land dedicated to energy crops used in
the model, as the energy crops production can only take place in a sustainable
manner.
Other sustainability constraints, beyond the ones set out by the RES Directive and the
Fuel Quality Directive or enhancement thereof can also be incorporated in the PRIMES
Biomass model, such as constraints concerning sustainable use of fertilizers and the
quality of water and air. Extensions of the sustainability criteria to imported fuels can
also be incorporated in different ways e.g. by assuming higher prices or reducing the
quantities available for imports.

Other policies The model can further implement other policies and measures. In different scenario
contexts, policies towards climate change mitigation can be simulated, such as policies
facilitating the use of Renewable Energy Sources and the application of carbon values
to the ETS and non-ETS sectors. Furthermore, measures such as subsidies can be
effectively incorporated in the model, as well as sensitivity analysis on the effect of
various parameters, such as conventional fuel prices, on the biomass supply system.

Q.5. Database of PRIMES biomass model

The construction of the biomass database was one of the most demanding and time
consuming tasks in the model development process. Extensive literature research has
been carried out in order to establish a reliable technical and economic database for
each stage of biomass conversion chain. The current model database has been
thoroughly updated in autumn 2011, when the technology and process data were
 updated. The historical data are updated to the latest available 2010 statistics. The
model databases were recently harmonised with other European models within the
context of the Biomass Futures project, while extensive use was also made of data
developed from previous projects, such as VIEWLS, REFUEL, BIOPOL and JRC.
The database of the PRIMES Biomass model has several components, which can
broadly be classified as: the historical statistical data; techno-economic data related to
technological parameters for the processes; country specific data relating to
agricultural/land use parameters as well as cost data; and import/export data
referring to the trade of commodities outside the EU.

Historical data The PRIMES biomass model uses historical data from 2000 to 2010 for calibration and
is able to represent the historical biomass situation in the EU. Where data is available
the model is, like all PRIMES family models, fully calibrated to Eurostat; as not all data
for biomass is available in Eurostat the model uses also further information sources,
such as FAOstat and Enerdata. The effort of collecting, analysing and filtering the most
reliable data available for the past years, demands a long time endeavour. This
process was fully concluded in autumn 2011.

Technical- The techno-economic data specifies the characteristics of the technologies and are
updated to reflect the latest technology developments; the projections for the
economic data development of technologies to the future were also updated to the latest available
data and literature available.
The data underlying the final values used within the PRIMES Biomass model are taken
from a large variety of sources including ECN and OEKO; expert judgement, external
consultation with experts, in particular when technology data was not found in
literature or when it was not possible to determine the robustness of a data source.
HISTORICAL DATA DESCRIPTION SOURCE
Bio-energy production Amounts of final bio-energy commodities Key source for the data concerning bio-
produced energy production for historical years is
Eurostat
Production technologies used The information concerning the
production technologies used derive
from several sources, such as Aebiom and
EurObserver
Land data Includes the cultivated land per crop for the Aebiom and other sources
production of biofuels for historical years.
Technical data for historical All techno-economical information needed for The techno-economical information used
years historical years, including costs per processes mainly came from ECN and OEKO and
(capital, fixed and variable costs), heat rate of were complemented by studies of NTUA
processes followed, fuel consumption and the Agricultural University of Athens
Energy crop production This data set includes all the essential The data concerning the production crop
cost information for the computation of the of the energy crops was derived from
production cost of energy crops for historical various sources such as USDA, FAO and
years per crop type and country (land yield, land FAOSTAT and several others were
renting cost, labour cost, cost of equipment, cost consulted such as the International
and uptake of fertilizers and nutrients, fuel Fertilizer Agency
prices)
Imports data Information regarding the trading activity that Data from various sources were used
took place internally in the European Union and including the NREAPs
among European Union and the rest of the
world.
Fuel prices Fuel prices for fossil fuels used during the PRIMES model: based on Eurostat and
production process Enerdata

TECHNICAL-ECONOMIC DESCRIPTION SOURCE

DATA

Cost per process Here capital, fixed and variable costs are
represented per process, from historical and
current years to future estimations of
technological maturity. For the collection of the technical-
economic data many sources and reports
Heat rate Model heat rate is used to indicate the efficiency were consulted. External consultation
of each process. with experts from the Chemical
Technical Lifetime Technical lifetime for every transformation Engineering Department of NTUA and the
process. Agricultural University of Athens took
place.
Amortisation The period to amortise a process investment.
The data that form the PRIMES Biomass
Utilisation Utilization rate of a production facility. model database are harmonised with
other European models.
Technical availability Estimates about the availability of a technology
at a commercially mature level
Fuel consumption Amount of energy consumed per technological
process.
 In PRIMES Biomass model, numerous conversion pathways are combined to shape
the biomass to energy conversion route, producing a variety of bio-energy products. A
schematic overview of the biomass conversion technologies effectively used in the
model is presented in Annex.
Data was researched for each component of the process in order to have updated data
for technologies, which currently do not exist, or for which data are not available, e.g.
data concerning the processes for the conversion of aquatic biomass to final bio-
energy commodities.

Country- This data refers mainly to agricultural and land use parameters and data referring to
costs, including cost supply curves for feedstock as well as commodity prices
specific data
(electricity, gas, other liquid fuels). The data mentioned in this section refers to
country specific data about future developments; past years are covered in the section
on historical/statistical data.
Several sources were used to construct the primary biomass potential databases. The
available energy crops production is determined endogenously by the model using
exogenous assumptions pertaining land availability and land productivity yields. The
yields are crop specific and are assumed to increase overtime due to technology
improvements in agriculture and additional agricultural policies. The model uses
curves to simulate different types of land with different land productivity and
fertiliser needs.
Concerning primary biomass potentials, several sources were used to form the model
databases. Information on energy crops were mainly derived from EEA studies and
EUWood data and estimates was used to determine forestry potential. The municipal
waste and landfill potential is based on values derived from GAINS and own analysis.
The analysis was based on the population growth estimations for each Member State
and used data derived from Eurostat waste statistics. Expert judgements were used in
order to disaggregate the waste potential derived from Eurostat into the four
categories used in waste management. Thus, the amount of waste land filled,
composted, incinerated and recycled was determined and therefore the waste
potentials that can be effectively used for energy purposes were specified. Regarding
black liquor, studies were used to determine current potential, whereas potential
projections to future years followed paper and pulp industry growth rates.
EU27 MEMBER STATE DATA DESCRIPTION SOURCE
Potentials Potentials for all biomass types of feedstock A number of sources were used for the
resources identified, are available, within the construction of the potentials database,
PRIMES biomass model in great detail. such as EUwood, EEA, Alterra etc.
Demand on biofuels and PRIMES biomass model is linked with The most frequently used data source
other bioenergy products PRIMES core model as it is determined to is the PRIMES Energy System Model.
compute all the outputs to meet a given Depending on the scenario other
demand of bioenergy products projected by external sources can be used, e.g. the
PRIMES model. The demand is provided by National Renewable Action Plans
country and by fuel. External sources which (NREAPs) submitted by the EU
give biofuel demand can be used Member States

Energy fuel prices Another input for PRIMES biomass model, PRIMES: the costs for the fuels depend
are the fuel prices of electricity, diesel oil and on the scenario context in which the
natural gas calculated by PRIMES core scenario is run
model, that are being consumed through the
various biomass transformation processes,
to produce final bioenergy products from
primary biomass feedstock.
Cost supply curves Estimated economic supply curves for all
biomass supply categories in which the
initial biomass primary resources have been
analysed.
Land data Land availability for dedicated energy crops
cultivation for every European Member
State.
Energy crops yield Possible yields for different kinds of energy
crops (sugar, starch, oil, wood lignocellulosic The data concerning energy crops was
and herbaceous lignocellulosic crops) derived from various sources such as
differentiated per European country taking EEA, USDA, FAOSTAT, EUWood, as well
into consideration climate and currently as from previous projects such as
dominant types of crops. VIEWLS and REFUEL
Energy crops production Detailed information on land renting cost,
cost land yield, labour cost, cultivation cost, price
and crop absorption factor of fertilizers and
nutrients and fuel prices for agriculture
(given by PRIMES core model), that are
available per energy crop type and European
country, result in calculating the overall
energy crop production cost.
GHG Emissions To compute the total CO2 emissions and PRIMES Energy System model
emission savings resulted from the extensive IPCC methodology for the calculation
use of biomass derived energy, emission of N2O emissions
factors from PRIMES core energy model for IFPRI, OEKO for ILUC emissions
electricity, diesel oil and natural gas are
included within the inputs of PRIMES
Biomass model. For electricity the values are
country specific based on the mix of fuels in
power generation and they change over the
years based on the scenario projection.
Moreover percentages that simulate the
abatement of CO2 emissions that needs to be
accomplished according to the EU
Renewable Energy Directive are included.
Trade & Import The Primes Biomass model allows trade of biomass feedstock and end bio-energy
commodities between Member States, as well as between the EU and the rest of the
data world. Information concerning the trading activity of Europe, both internal and
international, is covered. Historical data sets are collected from the year 2000 to 2010,
in order to comply with statistics. The necessary data for the construction of this part
of the database were derived from several sources.
TRADE & IMPORT DATA DESCRIPTION UPDATES THROUGH BIOMASS
FUTURES PROJECT
Potentials Potentials for all products (biomass
feedstock or end energy products)
imported internationally are available in
detail.
Data from various sources were
Imports exports supply curves Estimated economic supply curves for all
used to form this part of the
international imports and exports
database, such as IEA, Enerdata,
activities.
Eurostat, NREAPs, the U.S. DOE,
Distances and trade Trade matrix simulating distances and
FERN and FAOSTAT
connections trade connections between member
states and rest of the world.
Transport Costs and means used for transportation
regarding internal European trade and
international imports
Q.6. Biomass Conversion Chains

Bioethanol Production Chains

Bio-gasoline
production chains
Advanced Bio-
gasoline
production chains
Advanced bio-
gasoline
production chains
Biodiesel
production chains
Advanced
Biodiesel
production
chains
Advanced
Biodiesel
production chains
Bio-kerosene
production chains
Bio-kerosene
production chains
Bio-methanol Bio-DME
production chain production chain
Biogas production
chains
Biogas production
chains
Bio-methane
production chains
Bio-methane
production chains
Waste gas
production chains
Bio Heavy Fuel Oil
production chains
 Solid Biomass
production chains
Waste Solid
production chain
Bio-hydrogen
production chain
R.PRIMES New Fuels Model
A novel PRIMES model extension has been developed which aims at capturing the
market penetrations and future role of synthetic fuels, hydrogen, electricity, heat,
steam, chemical storage, carbon dioxide as a feedstock source, as well as the synergies
and competition between them. It will have a horizon up to 2070, similarly to the
updated main PRIMES model. The model represents the process flow with
engineering and economic details. It simulates hourly operation of a system with
electricity, hydrogen, gas, heat, steam and synthetic fuels in a synchronised way to be
able to analyse storage and finally the benefits deriving from sectoral integration.
The novel PRIMES model extension includes an aggregated representation of the
electricity, gas, heat and biomass models of PRIMES integrated into the process flow
modelling. It includes an aggregated way fuel choice in the demand sectors, but
demand for useful energy is exogenous coming from PRIMES.
It includes alternative pathways for the production of numerous low or zero-carbon
energy carriers, such as hydrogen, synthetic methane and synthetic liquid
hydrocarbons produced via Power to X (PtX) routes. At the same time, it includes
conventional energy carriers such as fossil hydrocarbons, biofuels, electricity, steam,
heat, etc. Given the large penetration of variable renewables in the future EU power
mix, the need for electricity storage will become more and more prominent. The new
module of PRIMES is operating at an hourly resolution and it can capture effectively
the operation of large-scale power storage systems. For example, the module
determines the hours of the day with excess renewables generation in order to
produce energy carriers (e.g. hydrogen) that can be used for the production of
electricity later when renewable generation is limited (storage). However, at the
same time, the new module is also able to decide whether economics favour the
production of synthetic hydrocarbons (using e.g. hydrogen as feedstock) instead of
providing storage services. In this way, it captures competition for carriers that can
serve different purposes (storage vs. feedstock) for different customers (power
generator vs. synthetic fuel factories).
Figure 6: A process flow diagram of the new PRIMES sub-model

All the aforementioned factors must be considered simultaneously, and along with the
operation of the rest of energy system (e.g. demand for synthetic kerosene,
availability of biofuels, etc.). Therefore, the new module includes aggregate
representations of the electricity, gas, heat and biomass models of PRIMES, integrated
into the process flow modelling, thus it includes, in an aggregated manner,
endogenous fuel choices in the demand sectors. Demand for useful energy is
exogenous coming from PRIMES. The remaining models of the PRIMES suite modules
are then calibrated so as to reproduce the fuel mixes as calculated by the new model.
E.g. PRIMES-TREMOVE respects the share of synthetic gasoline vs. bio-gasoline (and
petroleum-based gasoline) used by cars, as this is calculated by the new model
extension.
The module is pan-European and solves all countries of Europe simultaneously in
order to capture trade of the carriers, location of new factories and infrastructure
(power grids, gas, H2 network and distributed heat). It optimizes the investments and
operation of the system under perfect foresight assumptions. It includes several non-
linear mechanisms:
• Cost-potential non-linear curves for exhaustion of resources
(increasing slope)
• Non-linear learning curves (technology) and economies of scale
(factory), both with decreasing slopes
• Uncertainties and heterogeneity implying hidden-perceived costs
which non-linearly depend on enabling conditions (such as a carbon
price).
In a nutshell, the new model covers the following energy forms:

• Electricity: It can be produced via numerous sources, either fossil or

carbon-free. Electricity can be stored in a plethora of ways, either
 directly in batteries, or via the conversion to intermediate energy forms
(pumped storage, chemical storage as hydrogen, methane etc.).
• Heat and steam: Produced via heat pumps, boilers, CHPs units, for
distributed or on-site consumption.
• Carbon dioxide: Carbon dioxide serves an important role; it acts as the
main feedstock source for the production of synthetic hydrocarbons. It
can be captured from air or via applying CCU technologies to energy and
industrial applications. Only the former though guarantees that the
synthetic fuels produced will be carbon neutral (or even providing
negative emissions, in case they are combusted in biomass fuelled
power plants equipped with CCS technology-BECCS).
• Hydrogen: Carbon-free hydrogen is assumed to be produced via
electrolysis running on renewable electricity. It can serve as an energy
carrier (either combusted or used in fuel cells in stationary or mobile
applications), as feedstock for the production of synthetic fuels, or as a
means of storage for balancing the generation of variable renewables.
Hydrogen can be transferred via dedicated pipelines (that require
investments in infrastructure) or blended in the natural gas stream up
to a certain share (15%) due to technical limitations.
• Biofuels (liquid and gaseous) – They are produced using feedstock of
biomass origin. The model distinguishes fungible from non-fungible
biofuels. The former can fully substitute petroleum products, the latter
are blended up to certain shares with fossil based gasoline and diesel
because of technical limitations. Upgraded biogas (bio-methane) can be
blended to the natural gas stream.
• Synthetic methane – Synthetic CH4 is an output of a process such as
methanation, which utilises hydrogen and carbon dioxide as inputs. The
process is energy-intensive, requiring large amounts of electricity. This
carrier is usually referred to as “clean gas”, since its net carbon intensity
is lower than the one of natural gas, and the pathways to produce it as
Power-2-Gas (P2G, PtG). Depending on the origin of CO2 synthetic
methane can be considered even as carbon free, if the CO2 is captured
from ambient air.
• Synthetic liquid hydrocarbons – Usually referred to as Power-2-Liquids
(P2L, PtL); such fuels can fully substitute petroleum based products in
mobile applications with no radical changes in ICE powertrains. The
conventional powertrains continue to run on fuels with characteristics
similar to the ones of conventional oil products. Such vehicles exhibit no
range limitations and therefore synthetic fuels could be more easily
adapted by transport consumers. The competition with the
“electrification of the transport system” can hence be assessed by the
enhanced PRIMES model characteristics. However, PtL fuels would
probably find more room for development in transport modes, where
decarbonisation options are limited (aviation, long distances road
freight transportation), where they have to compete only with
advanced biofuels (with limited domestic resources) and/or
technologies that are currently at low TRL levels (e.g. electric aircrafts).
Synthetic hydrocarbons are produced in the model with two main
 pathways whose intermediate products are either syngas (blend of CO
and H2), or alcohols (methanol).
• Electricity Storage – Although technically, not an energy form, power
storage is an important element of the new module. Storage can be
served either via batteries or via the intermediate step of producing
hydrogen and/or synthetic methane for later use. The operation of the
module at an hourly resolution allows capturing the appropriate time
segments for power injection to storage, and extraction from storage at
a later time interval.
• Fossil fuels – serving as conventional energy carriers. Their use results
in GHG emissions.
The graphic below illustrates the linkage of the new fuels model with the chemicals
(fertilizers and petrochemicals) structure, to possibly replacing naphtha and natural gas
reforming by RES-generated liquid fuels. This would be a way of storing carbon dioxide
into materials. In this way, the PRIMES model handles the relationship between the
industrial energy model and the industrial processing model regarding the possibility of
reusing captured carbon dioxide.
The power model also handles several options of carbon dioxide capturing: a) CCS and
CCU from fossil fuel plants, b) From biomass plants equipped with carbon dioxide
capturing, c) from the ambient air, d) From the industrial sector capturing for reuse. The
model ensures a balance between carbon dioxide capturing, reuse and storage in
materials or in geological caverns. The optimisation is cost-based and takes into account
nonlinear cost-supply (potential) curves of geological storage, that represent social
acceptance as hidden cost factors.
 S. Greenhouse Gas Emissions and Policies
S.1. Emissions
CO2 emissions from energy are computed by multiplying quantities of fossil fuel
combusted (measures in energy terms) by emission factors, which are specific to
fuels.17 Abatement of energy-related CO2 emissions is an endogenous result of the
model and depends on fuel mix, technology and process mix within the energy
system. The model projects a configuration of the energy system to the future and
computes energy-related CO2 emissions. The model does not include any explicit
marginal abatement cost curve (MAC) for energy-related CO2 emissions. Abatement
and its relation with costs result from energy system simulation. MACs can be
quantified using the results of the model for a variety of imposed abatement levels.
When in a scenario CO2 pricing, taxation or quantity limitations apply, the model acts
from emissions to the energy system simulating a feedback effect. A quantity
limitation on emissions may be treated at the level of each sector or for a country’s
energy system or for the EU as a whole. In such cases, the model also considers the
shadow value of the carbon constraint, which is termed carbon value and influences
demand and supply decisions of agents.
A carbon value differs from carbon taxes as it does not entail direct payments,
although may inducing higher indirect costs. When policy instruments involving CO2
emission allowances apply, the quantity of available emission allowances are
represented as a carbon constraint and the shadow value, which is computed by the
model, acts as a carbon value. When in addition allowances are purchased on
auctions, then direct payments are induced and the model considers them in the
simulation of agents’ choices.
CO2 from process emissions are computed through simple relationships which involve
physical production of the relevant industrial commodities (e.g. cement). Simple
techniques that may reduce such emissions are represented through marginal
abatement cost curves. Higher emission reductions are represented by assuming
Carbon Capture and Storage techniques, which apply on the processing of industrial
commodities. The representation includes capital and variable costs of CCS, as well as
electricity consumption associated with capture, which adds up to total demand for
electricity.
PRIMES covers Emissions of non CO2 Greenhouse Gases are included in the PRIMES report based on
CO2 emissions calculations using marginal abatement cost curves and projections quantified by the
from energy and
industrial GAINS model of IIASA.
processes and in
coordinated use When in a scenario GHG emission constraints apply, the PRIMES model computes
with GAINS all
GHG emissions.. allocation of reduction efforts to the various emission components (energy CO2,
process CO2 and the various non CO2 GHGs) by considering equalisation of carbon
values (i.e. equal marginal abatement costs). Sector specific emission reduction
constraints can also be treated.

S.2. Emission Reduction

The approach for modelling EU policies concerning greenhouse-gas-emission
reduction includes the following:

17Emission factors by fuel are in accordance to 2007/589/EC. In specific cases where the country specific
emissions diverge substantially from the emission factors of 2007/589/EC, country specific emission
factors are used.
 • The model can analyse various emission constraints: per sector, per country
or EU-wide
• The sectors are grouped in ETS and non ETS, with different representations of
mechanisms
For ETS
• An EU-wide emission constraint is applied reflecting total volume of
allowances (per year) and assumptions about permissible international
credits (e.g. CDM)
• Grandfathering (free allowances) can be represented through exogenous
quotas per sector and per country; carbon prices are, entirely or partially
(reflecting degree of market competition), treated as opportunity costs and
price signals, but actual payments only correspond to excess emissions by
sector
• Auctioning of allowances is represented by modelling carbon prices inducing
true payments by sector
• Carbon prices are determined iteratively (until ETS volume of allowances is
exactly met) and apply on all ETS sectors and countries in a uniform way
• Inter-temporal aspects, such as arbitraging over time within the ETS, are
considered in the modelling by introducing cumulative allowances as a
constraint and excluding borrowing from the future (the model running is
however iterative, as inter-temporal optimisation was not technically possible
because of computer limitations)
For non ETS
• The model can handle non ETS emission reduction targets either on a country
level or EU-wide assuming possible exchanges between MS
• Carbon values (i.e. shadow prices associated with the volume constraint)
serve to convey price signals to non-ETS sectors without entailing direct
payments (only indirect costs)
• Carbon prices and carbon values act on top of any other policy measure (of
specific character, for example standards, specific taxes, subsidies, RES
policies and obligations, etc.), thus ETS carbon prices determined
endogenously depend on the extend of other policies and measures assumed
for a scenario
LULUCF emissions
• Included through linkage with models GLOBIOM and CAPRI, which take inputs
from PRIMES biomass model
PRIMES does not model the international market for carbon credits (e.g. CDM).
Usually the scenarios assume that the EU is a price-taker of the marginal CDM price
and that there is an upper bound on the volume of carbon credits to be taken from
CDM. Thus, if the assumed CDM price is lower than the estimated EU ETS carbon
price, carbon credits from CDM are taken up to the upper bound. Using PRIMES in
linked form with Prometheus model or other global model (e.g. GEM-E3, POLES), it is
possible to simulate global ETS and carbon markets with different groupings
(bubbles).

S.3. ETS Market Simulation

ETS Market From a modelling perspective it is complex to optimise dynamically in order to
achieve emission targets which are defined in cumulative terms. This is the case of the
Simulation EU ETS regulation: EU ETS allowances are to be decreased by a fixed amount every
year calculated by applying 1.74% on base year ETS emissions and allowances not
used to justify emissions can be banked for future use but borrowing from allowances
to be issued in the future is not permitted.
Cumulative emission constraints (carbon budget) are also included as an emission
constraint in decarbonisation scenarios handled using PRIMES. In such cases, the
modelling is complex because of dynamic arbitration due to the cumulative character
of the carbon budget, which must be met. A usual policy request is also to include
emission reduction targets for specific years (i.e. 2020, 2030 and 2050) or for one
period (i.e. 2030) together with a cumulative carbon budget for the entire projection
period.
Mathematically, determining ETS carbon prices as a function of time under
constraints about cumulative emissions and end-horizon emissions is a typical
optimal control problem, which resembles to optimal control problems applied in
economics of depletable resources. Take the example of oil extraction: in the analogy,
carbon prices are the oil prices, cumulative emissions are the oil reserves and the
annual abatement cost curve is the annual oil production cost curve. The result of this
problem is that if marginal costs of production (emission abatement) remain
unchanged over time then the rate of exhaustion of reserves (the annual emissions)
must be such that the resulting annual price of oil (carbon price) is exactly equal to
the discount rate used to calculate the present value of revenues and costs (auction
payments). This is called a Hoteling rule for price determination.
From a computational perspective, it is impossible to solve the entire PRIMES model
as an optimal control problem because this would imply full inter-temporal balancing
of all demand and supply optimisations simultaneously with the inter-temporal
clearing of the ETS market. The computer time would be very large.
Instead, PRIMES follows an iterative process. The demand and supply sub-models
apply inter-temporal foresight at various degrees (shorter in demand, longer in
supply) and so by anticipated ETS carbon prices influence agents’ decisions. Based on
expert judgment the model starts using a first approximation of time path of ETS
carbon prices. Then the loops of the PRIMES model run to find market equilibrium
until the end of the projection horizon. The resulting cumulative emissions and point
emissions (e.g. in 2020, 2030 and 2050) compare to targets. Depending on deviation
from targets, the time path of ETS carbon prices re-adjusts. The adjustment depends
on surplus of allowances. High surplus implies that holders of allowances tend to sell
as they anticipate waiting long time before obtaining the Hotelling-based return.
Depending on assumptions about their foresight horizon, they may behave more or
less risk-averse, which influences the timing of selling when high surpluses persist.
Agents who use allowances to justify emissions are also risk averse in situations of
high surpluses because of perception of high regulatory uncertainty when the market
is destabilised by high surpluses. Therefore, these agents tend to be reluctant in
banking allowances in periods of low prices, when high surpluses persist. In
summary, the ETS trajectory adjustment depends on surpluses.
Version 6 includes representation of the Market Stability Reserve (MSR) regulations
and the retirement of allowances in early stages. The MSR aims at reducing and
maintaining EUA surplus within boundaries: Emission distribution will reduce if
surplus >833MtCO2; Emission distribution will increase if surplus <400MtCO2. The
model’s logic on MSR is that the structural reform of the EU ETS reduces the risk of
holding banked allowances by those being long and provides an incentive to those
being short (such as power generators) to buy allowances at periods of low prices. As
the system is automatic the stability of the ETS system will be granted under all
circumstances (e.g. economic crisis or growth, strong change in international fuel
prices, changes in policies), providing for regulatory security and higher
predictability. Therefore, the structural reform will help smoothing the trajectory of
carbon prices.
If carbon values apply to non-ETS sectors and if they are equal to ETS carbon prices,
then carbon values also change in each iteration. It is reminded that each iteration
involves full running of PRIMES over the entire period, therefore the iterative process
is time consuming and includes human time spent to decide on the re-adjustment of
carbon price trajectory in each iteration.
T.Prices of Energy Commodities
T.1. Introduction
PRIMES takes as input projection of prices of imported fossil fuels in Europe. Usually
they are based on projections by world energy models such as Prometheus (E3-
Modelling), POLES (Grenoble or JRC) and IEA’s WEO. The international fossil fuel
price projections apply uniformly on all EU MS, but border prices are further
differentiated by country taking into account transport costs and systematic
differences of country prices from average EU import prices due to market reasons.
Such differences are calculated using observed time series of prices on which an
econometric function is possibly estimated serving to extrapolate the differences to
the future. It is taken care to avoid extrapolating price differences, which are due to
market distortions that may have occurred in the past.
The prices of domestically produced fossil fuel prices are based on costs of extraction
plus subsidies. The subsidies are either direct (known based on collected information)
or indirect, when for example prices of domestic fossil fuels are defines equal to
imported fossil fuel prices although domestic extraction cost is higher. Data are
generally missing to estimate the subsidy component in detail, so in many cases
PRIMES uses imported prices as estimates of prices of domestically extracted fossils.
Nonetheless, PRIMES has devoted much effort to estimate cost of extraction for coal
and lignite, by collecting data from national sources. In particular for lignite, PRIMES
include differentiated extraction costs by country, based on collected information.
Based on the above, PRIMES calculates fossil fuel prices at primary energy level as
weighted sum of prices at imports and costs or prices of domestic extraction. The
model projects fossil fuel prices in the domestic markets by sector of consumption, to
take into account price differentiation by sector. Pre-tax fossil fuel prices by sector are
calculated by adding three components, as follows. After-tax prices add excise taxes
and VAT rates as exogenous parameters.
a) cost corresponding to the primary energy price of the fuel,
b) cost of fuel processing (e.g. refinery, briquetting, coking, etc.) which is specific
to each fossil fuel type,
c) cost of transportation and distribution which is specific to the customer
(sector) type.
The latter component is supposed to reflect not only transportation and distribution
costs but also costs due to diseconomies of scale for small-scale customers (e.g.
delivery of briquettes to residential sector). The same component includes cost of
refueling at service stations for conventional transport sector fuels.
For natural gas, power generation is supplied by high pressure network, industry by
high pressure or medium pressure pipelines and residential and services customers
by low-pressure pipelines. The regulated tariffs of transport and distribution are
different for each type of pipeline, hence by sector. Similarly the requirements for gas
balancing are different by sector, thus prices of gas differ by sector, also because of
weighting pipeline and LNG gas or services from storage. Gas balancing costs are
estimated using shares of LNG and gas storage in gas supply. The regulated tariffs are
calculated and projected by the model, based on costs of regulated asset basis and on
a levelized costing method using a regulated discount rate. The tariff computation is
compared and validated with information collected for past years.
The prices of bio-energy commodities are projected using the PRIMES biomass model
which includes a pricing model. Similarly, electricity and distributed steam/heat
prices are calculated using the PRIMES power/steam model, which includes a pricing
model. The unit costs of refinery products are computed using the refinery module.

T.2. Mathematical illustration of price calculation

The methodology described below applies to crude oil, refinery products, natural gas,
coal, lignite, coke and biomass or waste products.
The prices of imported fossil fuels (𝑤𝑗 ) are exogenous, namely for crude oil, natural
gas and coal. The unit costs of domestically produced fossil fuels are also exogenous
(𝑑𝑗 ).

The price 𝑝𝑗𝑖 of a commodity 𝑗 paid by a consumption sector 𝑖 considers as separate

components the price of energy 𝑚𝑗𝑖 and the tariffs paid for transport and
distribution 𝜑𝑗𝑖 . Additive taxes or subsidies (𝜏𝑗𝑖 ) or multiplicative taxes (𝜎𝑗𝑖 ), where
applicable, are exogenous.
The general formulas for computing the price are as follows:

𝑝𝑗𝑖 = (𝑚𝑗𝑖 + 𝜏𝑗𝑖 ) ∙ 𝜎𝑗𝑖 + 𝜑𝑗𝑖

𝑚𝑗𝑖 = 𝛼𝑗𝑖 + 𝛽𝑗𝑖 ∙ [𝜆𝑗𝑖 ∙ 𝑤𝑗 + (1 − 𝜆𝑗𝑖 ) ∙ 𝑑𝑗 ]

𝑜𝑟 ∆ ln 𝑚𝑗𝑖 = 𝛼𝑗𝑖 + 𝛽𝑗𝑖 ∙ ∆ ln[𝜆𝑗𝑖 ∙ 𝑤𝑗 + (1 − 𝜆𝑗𝑖 ) ∙ 𝑑𝑗 ]

−𝑡
𝑅𝐴𝐵𝑗 + ∑𝑡 𝐶𝑗,𝑡 (1 + 𝜌𝑗 )
𝜑𝑗𝑖 = 𝜃𝑗𝑖 ∙ −𝑡
∑𝑡 𝐷𝑗,𝑡 (1 + 𝜌𝑗 )

In the above, 𝜆𝑗𝑖 represents the main driver of prices and usually, but not necessarily,
it takes values 0 (domestic costs prevail) or 1 (import prices prevail). It is possible,
that import prices are drivers of a commodity price, although the commodity is
domestically produced; in this case opportunity costs drive pricing. The additive
factor 𝛼𝑗𝑖 results from two components, 𝛼𝑗𝑖 = 𝛾𝑗 + 𝛿𝑗𝑖 , where 𝛾𝑗 measures supply costs
to a country which apply in addition to import prices (or extraction cost) to reflect
transport costs, special contracting conditions, etc., and 𝛿𝑗𝑖 is an additive component
of the pricing conditions for sector 𝑖, as for example volume discounts, supply costs,
etc. The multiplicative factor 𝛽𝑗𝑖 represents the degree of correlation of commodity
prices with import prices or domestic costs.
When information exists allowing decomposition of imports of commodity 𝑗 by origin,
the equation for 𝑚𝑗𝑖 includes 𝑤𝑗 also decomposed by origin. For example, consider
natural gas imports decomposed in pipeline gas and LNG. Different price components
form 𝑤𝑗 and the weight of LNG can be differentiated by sector 𝑖 to represent
differentiated use of balancing gas in relation to the demand profile of the sector.
The tariffs of transport and distribution apply only for network-based carriers. Their
computation uses the regulated asset basis,𝑅𝐴𝐵𝑗 , which is estimated based on
network length and characteristics, the anticipated investment and maintenance or
operation expenditures in the future, 𝐶𝑗,𝑡 , the anticipated demand, 𝐷𝑗,𝑡 , and the
regulated discount rate 𝜌𝑗 .

The parameters in the formulas above are computed in the calibration stage of the
modelling, so as to reproduce base-year prices. The parameters 𝛼𝑗𝑖 and 𝛽𝑗𝑖 are
computed for refinery products by the detailed refinery sector module of PRIMES,
only if the model running has included the refinery module, otherwise the
computation uses calibrated coefficients. The detailed biomass supply model of
PRIMES computes the domestic component of the prices of biomass or waste. For the
supply of heat by district heating networks, the model computes distribution costs as
shown above and uses estimation of average cost of heat production, as the domestic
cost component 𝑑𝑗 , resulting from the power and heat model of PRIMES. The unit
domestic cost 𝑑𝑗 includes any environment-related cost or tax item that possibly
applies on domestic production.
U. PRIMES reporting on Energy System Costs
PRIMES reports on costs and prices by sector in detail.
For policy evaluation, PRIMES reports on costs from the perspective of final energy
consumers, namely industry, households, services and transportation. Such costs per
sector are decomposed in:

 Annuity payments for capital based on the sector’s discount rate (alternatively
annualised cash payments for investment)

 Annuity payments for direct energy efficiency investments (or annualised

cash payments for investment)

 Variable costs for operation and maintenance

 Fuel, electricity and distributed steam/heat purchasing costs (which reflect all
costs incurring by energy suppliers, including taxes, ETS, etc.)

 Direct tax payments

 Disutility costs (income compensating variation of utility applicable for

residential, services and transport of individuals)
Adding these costs for all demand sectors the model computes the Total Energy
System Costs, which can be seen as payment by the rest of the economy in order to get
the required energy services. Obviously the total energy system costs does not refer
only to the purchasing of energy commodities but also to all kinds of expenditures
incurring to consumers for energy purposes, since equipment purchasing costs and
energy efficiency enabling expenditures are included. Total energy system costs are
also reported as % of GDP to indicate the cost of energy services for the economy.
Auction (ETS) revenues may be excluded from total energy system cost when
assuming that the recycling of public revenues in the economy is performed without
transaction costs. The reported total energy system cost excluding auction payments
as % of GDP is thus a better measurement of energy cost impacts from a
macroeconomic perspective.
Tax revenues from energy are reported separately include revenues from excise taxes,
VAT rates and ETS auctions.
Cash payment for investment is also reported by sector and by type of investment,
separately including investment in infrastructure.
The expenditures in energy-related equipment, such as for appliances and vehicles,
are calculated based on total purchasing costs, which of course does not correspond
only to the energy-related component of the equipment. To isolate energy-related
costs, PRIMES calculates an indicator of incremental capital costs specifically for
transport equipment: incremental unit capital costs (per vehicle) in a future year
relative to base year values are assumed to occur for energy and emission reduction
purposes and so the increment is used to calculate capital costs entering Total Energy
System Costs.
V. Methodology on Discount Rates
V.1. Overview of discount rates within a modelling approach
The PRIMES model explicitly considers the time dimension and performs dynamic
projections. The model projects decisions in which they explicitly consider the time
dimension of money flows as perceived by actors. Following microeconomic theory,
they actors have preferences about the time dimension of revenues and costs, in the
sense that they have to discount an amount defined at future time to make it
equivalent to an amount available at present time. The time preference18 has nothing
to do with inflation and is subjective.
The PRIMES model mimics decentralised decisions of the actors so that each actor can
apply his individual discount factor, in contrast with other models, which formulate
central planning optimisation and assume that the central planner applies a uniform
discount factor on behalf of all actors.
The central planning approach can be characterised as normative, whereas the
descriptive approaches, as PRIMES follows, use market-based discount factors
differing by agent. PRIMES follows a descriptive approach because it aims at assessing
policy impacts as close as possible to reality in order to avoid under- or over-
estimation of the costs and difficulties of transformation towards meeting targets.
Central planning models may have different aims, as for example to evaluate what
should be the “optimum” system if the world was ideal. They project technology
diffusion in the context of an idealised world, which can be misleading for policy
making aiming at promoting technology diffusion.
Capital-budgeting decisions enter the PRIMES model in all sectors, both in demand
and supply of energy. The simulation mimics the appraisal undertaken by a decision-
maker of whether purchasing of equipment or investing in energy savings or
infrastructure is worth the funding. The decision involves comparison among
alternative options, e.g. technologies, which have different proportions of upfront
costs and variable operating expenditures (including fuel costs). As the cost structure,
in terms of CAPEX and OPEX, differ across the various options, the decision maker has
to do arbitration over time. Therefore, the decision maker’s time preferences, in other
words the discount factor, influences his choices. The time preference is inherently
subjective and the decision maker appraises whether the upfront spending is worth
the funding, compared to other options of using the funds, while taking into account
uncertainty surrounding the investment options and the scarcity of funding.
Therefore, the value of the discount factor depends on many factors, such as the
interest rates prevailing in capital markets, the degree of access to such markets for
fund raising, and mostly by the value that the actor associates to own funding
resources, such as equity capital or savings of individuals.
Therefore, private discount factors reflect opportunity costs of raising funds by the
actor on a private basis. Obviously, the opportunity costs of raising funds differ by
sector and by type of actor, being very different by income class. They also vary with

18 In economics, time preference is the relative valuation placed on a good at an earlier date compared with

its valuation at a later date. In mathematical terms, the decision maker uses a discount factor, say 𝑑 (a rate
measured as a percentage), so as to be indifferent when has to choose between a present amount 𝐹 and a
future amount 𝐹 ∙ (1 + 𝑑)−𝑡 available with certainty time t.
the degree of risk associated to the decision options. In contrast, social discount
rates19 reflect opportunity costs of raising funds by the state or the society.
Private discount factors depend on policies when for example actors use high
discount rates due to market distortions and non-market barriers. There are many
examples of policies influencing discount rates in sectors such as energy efficiency,
renewables and even nuclear or CCS investment. The state may apply support
schemes to mitigate risks and reduce the individual discount rates, such as feed-in-
tariffs (FIT), contracts for differences (CfD), power purchase agreements (PPA),
sovereign guarantees on investment, reduced taxation, subsidies on interest rates,
and generally innovative financing mechanisms. Policies may also transfer risk
hedging from individuals to institutions, the latter being able to manage risk
collectively and thus more efficiently; examples are the energy service companies
(ESCO), the policies obliging utilities to save energy at the premises of their
customers, the loans by development banks, etc. All these policies show in PRIMES as
reductions of individual discount factors.

V.2. Capital budgeting decisions in PRIMES

An investment choice always involve upfront costs and variable-operating
expenditures or revenues which take place over time (e.g. annually). The decision
draws on a comparison of different investment options.
The PRIMES model uses different capital budgeting methods in the various sub-
models. Examples are as follows:

 In the standard version of the power sector model, the choice of options for power
capacity expansion uses equivalent annuity costs (EAC). This is part of
intertemporal minimization of costs, which guide investment choices within
stylised generator portfolios. In the model version, which represents market
imperfections, the calculation evaluates expected Net Present Value of investment
(NPV), which include risk aversion factors, for each capacity expansion option.
The actor either invests and chooses an option or decides not to invest at all.

 In the sub-model, which calculates, investment based on feed-in tariffs or on

contracts for differences (CfDs) the model uses a method based on Internal Rate of
Return (IRR) calculation by type of investment project from which it derives the
probability of investment implementation. Instead of assuming a single threshold
value for acceptable IRR, the model uses a frequency distribution of threshold
values depending on the IRRs in order to capture heterogeneity of actors and
different investment circumstances.

 In the sub-models, which calculate tariffs for using infrastructure subject to

regulation as a natural monopoly (power grids, gas network, recharging
infrastructure for vehicles, etc.), PRIMES follows the NPV method and uses the
regulated rate of return as discount factor.

 In the sub-models, which include investment options for energy savings (e.g.
insulation of buildings, control systems in industry, etc.) PRIMES calculates
equivalent annuity costs of the energy saving investment and compares annual
capital costs to economised annual expenditures due to lower energy
consumption. The model calculates a payback period, which combines with
frequency distribution of threshold values reflecting heterogeneity of consumers
and installations, to determine likelihood of investment.

19If social discount rates are used in simulations of private investment decisions, the modeller implicitly
assumes that the economy has no funding scarcity and perfect capital markets allow unlimited liquidity.
In the demand sub-models, which include technology choice by type of equipment or
vehicle, the formulations calculate equivalent annuity costs for each option and
formulate a frequency distribution of technology choices based on relative EACs to
reflect heterogeneity of consumers.

V.3. Methodology for defining values of discount rates

The model follows different approaches by sector.

V.3.a. Decisions by firms generally follow the approach of the weighted

average cost of capital (WACC) to define discount rates.
The WACC expresses the unit cost of capital for a firm depending on the source of
funding, with each type of source using a different interest/discount rate. The main
distinction is between equity capital (𝐸) and borrowed capital (𝐷). The former is
valued at a subjective discount rate 𝑟𝑒 and the latter at a market-based lending rate 𝑟𝑑 .
A simple WACC formula is as follows:
𝐸 𝐷
𝑊𝐴𝐶𝐶 = 𝑟𝑒 + 𝑟
𝐸+𝐷 𝐸+𝐷 𝑑
To determine the discount rate on equity the model follows the methodology of the
capital asset pricing method (CAPM) which is:
𝑅𝑒 − 𝑅𝑓
𝑅𝑒 = 𝑅𝑓 + 𝛽 ∙ (𝑅𝑚 − 𝑅𝑓 ) ⟺ 𝛽 =
𝑅𝑚 − 𝑅𝑓

𝑅𝑓 is the risk-free interest rate. 𝑅𝑚 is the benchmark or specific market rate of return
on capital (expressing the usual practice of the sector). 𝛽 is a subjective ratio
expressing risk premium of equity relative to risk free options over the usual risk
premium of the sector expressed by the difference of the market specific rate and the
risk-free rate. Obviously 𝛽 > 1 indicates a risk averse behaviour, which implies high
WACC values compared to risk prone behaviours using 𝛽 < 1. Technology- or project-
specific risk premium values correspond to using a value of 𝛽 higher than one.
An alternative formulation for estimating the unit capital cost of equity (COE) is to
decompose 𝑅𝑒 as follows:
𝐶𝑂𝐸 = 𝑅𝑒 = 𝑅𝑓 + 𝐸𝑅𝑃 + 𝑆𝑃 + 𝐼𝑅𝑃 + 𝐶𝑆𝑅𝑃

In the above 𝑅𝑓 is the risk-free rate, ERP the equity risk premium, SP the size risk
premium, IRP the industry risk premium and CSRP the company-specific risk
premium.
Surveys of equity risk premium indicate that the values used in practice differ by
country and over time reflecting country-specific and economic context-specific risks.
The valuation of country-specific or condition-specific equity risk premium is
relatively independent of the financial situation of the banking system concerning the
lending interest rates or the liquidity of funding by banks. Even under perfect bank
liquidity and low lending interest rates, equity risk premium can be high reflecting
country-specific uncertainties.
The literature has extensively published survey results of WACC rates (or the ERP
rates) used as common business practice in various sectors. The surveys show that
capital intensive sectors generally use lower WACC (or ERP) rates than labour-
intensive sectors; also WACC rates are higher in small scale projects compared to
large scale ones and the WACC rates can be significantly high in technologically
emerging sectors or applications. The WACC rates are lower for firms holding
dominant positions in markets. They are lower also when they are state-owned or
supported by the state, compared to firms operating in fierce competition conditions.
New entrant firm in a market with a dominant incumbent firm usually applies a
higher WACC than the incumbent, and this is part of the cost of new entry.
Based on these considerations, the PRIMES model applies (or can apply) slightly
different WACC rates by business sector, by type of technology (mature versus
emerging) and by scale level (e.g. industrial or decentralised versus utility scale).

V.3.b. Decisions by individuals using a subjective discount rate to

annualize investment (upfront) costs following the equivalent annuity
cost method.
A vast literature collected as part of PRIMES modelling research has published
numerous statistical surveys, which estimate the subjective discount rate that
individuals implicitly use when making a choice between equipment varieties having
different upfront costs and different variable operating costs.
A pioneering research20, back in the ‘70s, has used a large sample of data based on
surveys of purchasing of air-conditioning systems by individuals; the sample included
a variety of air conditioning types with different purchasing costs and different energy
efficiency rates. Using the sample, the author econometrically estimated the median
value of the discount rate that implicitly individuals use to make their choices. He
founds a median value between 24 and 26% for the discount rate and points out to
the fact that this value substantially exceeds values used in engineering calculations to
determine the so-called life-cycle costs for evaluating the trade-off between energy
efficiency and higher initial capital costs.
The low rates used in engineering calculations suffer from two shortcomings: from a
positive standpoint, they are too low to forecast accurately consumer behaviour and
thus can be misleading for policymaking purposes, while from a normative standpoint
they are too low to suggest how individuals should make their choice of equipment.
The lower bound of the individual discount rate (within the confidence interval based
on the sample population) is equal to 15%, which is also much higher than values
used in engineering calculations. The author compares the estimated values to the
interest rate of 18% applied on credit cards at that time and finds logical that
individuals value cash scarcity (opportunity costs of raising funding from a private
perspective) at a rate above the rate prevailing in the credit market. The state policy
may see the difference between the individual and the social discount rates as a non-

20Probably the first paper of this kind was the one by Jerry A. Hausman, Professor at MIT, USA Boston,
paper published as “Individual discount rates and the purchase and utilization of energy-using durables”,
The Bell Journal of Economics (Vol. 10, No 1, spring issue), 1979.
price market barrier, a sort of market imperfection. Therefore, policies based on
efficiency standards are effective for inciting energy-efficient choice of appliances in
circumstances with strong barriers compared to price-based policies, precisely to
offset barriers causing high individual discount rates.
The econometric analysis has also verified that the implicit discount rate has a
negative strong correlation with income and is as low as 3.6% for high-income classes.
Economic theory suggests that discount rates should decrease as income rises, even
with perfect capital markets, since the marginal income tax rate rises with income and
the gains from using efficient appliances are untaxed. A histogram of individual
discount rates depending on income level can be seen in the graph below. The median
value of the discount rates is 24% and the income elasticity is -1.5, which indicate a
remarkably high increase of the discount rate for low-income percentiles.
The findings mentioned above are common to numerous studies and publications
surveying purchasing behaviours for a large variety of equipment types. To illustrate
these findings, many authors proposed terms such as “energy efficiency gap” or
“energy efficiency paradox” to describe the implications of using high individual
discount rates rather than engineering-oriented or social ones.
Kenneth Train21, as well as Sanstad, Blumstein and Stoft22 summarised the findings of
many surveys of the ‘80s and ‘90s of consumer behaviour for a large number of
equipment. All surveys confirmed the strong inverse correlation of individual
discount rates and income. The estimations confirmed the large variation of individual
discount rates mainly as inverse function of income per household:

 14% - 56% for heating equipment

 5%-90% for cooling equipment
 5%-30% for automobiles
 4%-88% for insulation of houses
 15%-45% for double glazing and other similar measures in buildings
 15%-62% for cooking and water heating equipment
 4%-51% for boilers (difference with heating equipment, see first bullet)
 35%-100% for refrigerators and
 20%-40% for small black appliances.
The surveys23 also revealed that beside income, which is the main explanatory factor
of variance of discount rates, the range depends on the age of the persons and the
ownership of the property.
A similar approach uses the concept of hurdle rates, which express the minimum rate
of return on a project or investment required by the decision maker to compensate
for risk associated to future gains. Several econometric studies based on surveys
provided evidence that hurdle rates effectively used by individuals and small firms to
make investment decisions on energy efficiency are set at levels much above interest
rates considered by large firms for equity capital in the context of capital asset pricing
methods.

21 “Discount rates in consumers’ energy-related decisions: a review of the literature”, Energy, Vol. 10, No
12, pp. 1243-1253, 1985
22 “How high are option values in energy-efficiency investment?” Energy Policy, Vol. 23, Mo 9, pp. 739-743,

1995
23 The following references include data from surveys and econometric estimations of individual discount

rates: 7, 8, 11, 13, 18, 20, 21, 23, 26, 31, 37, 38, 39, 44, 55 (See references section)
 Modern behavioural economics propose models which deviate from classical
microeconomics (e.g. bounded rationality model24, loss aversion model25) which are
asserted to explain the persistence of high hurdle rates (equivalently discount rates)
in choices for energy-efficiency investments, with initial investments being given
asymmetrically greater weight than future savings.
Illustration of dependence of individual discount rates on income

Individual discount rates as function of income class 30%

120 25%

Frequencies
20%
100
Discount rates (%)

y = 107.33x-1.515
R² = 0.9422 15%
80
10%
60
5%
40
0%
20
0
<6000 $ 6000 - 10000 - 15000 - 25000 - >35000$
Individual discount rates
10000$ 15000$ 25000$ 35000$
But, despite the different explanatory approaches there is no doubt in the literature
about the persistence of high hurdle and discount rates at levels much above
engineering and social rates, as well as the strong inverse correlation of the rates with
income. Until today, there has been no statistical survey finding low hurdle or
discount rates for individuals making selection of energy efficient investment or
equipment purchasing despite the campaigns and strong policies favouring energy
efficient choices.
This advocates in favour of maintaining high values of discount and hurdle rates for
individuals consistently in the modelling. It is equally important to recognise that
neglecting heterogeneity of consumers (for example when modelling only a
representative consumer by sector) is a serious shortcoming of the modelling. This
call upon introducing frequency distributions by income class, or by size and other
characteristics and vary the discount rates across classes.
Based on the empirical findings about high discount rates, the literature proposed
many explanations. Overall, it is common that energy-efficient technologies entail
longer payback periods and greater risks and uncertainties than conventional
technologies. According to the reviewed literature, specific causes may include:

 lack of information about cost and benefits of efficiency improvements;

 lack of knowledge about how to use available information;

 uncertainties about the technical performance of investments;

24 Bounded rationality is the idea in decision-making, rationality of individuals is limited by the information

they have, the cognitive limitations of their minds and the finite time they have to make a decision.
According to this theory, the decision maker is a satisfier, seeking a satisfactory solution rather than the
optimal one. Nested decision making models, in which the first level nests refer to seemingly non-economic
choices (e.g. colour, convenience, and modernity) imply biased selection of lower level nests, which involve
economic considerations and thus the selection can deviate from economic optimality.
25 In economics and decision theory, loss aversion refers to people's tendency to strongly prefer avoiding

losses to acquiring gains. Most studies suggest that losses are twice as powerful, psychologically, as gains.
This point of view can be represented also by classical microeconomic theory by assuming strong risk
aversion.
  lack of sufficient capital to purchase more expensive but efficient products (or
capital market imperfections);

 income level and consequently savings resources;

 high transaction costs for obtaining reliable information;

 risk averse attitudes associated with possible financial failure of the

investment, etc.
Ownership status is a relevant socio-economic explanation of high implicit discount
rates.
The literature argues that the relationship between low income and high implicit
discount rates derive partly from poor access of low-income households to capital
markets and low liquid capital availability than higher-income households. As a result,
even when adequate information on investment returns are certain, lower-income
households are reluctant to invest in efficient technologies unless complementary
economic instruments are in place. All these explanations imply that appropriately
targeted policies have to be in place to reduce the individual discount rates and to
make energy efficiency investment more attractive to consumers.
The above-mentioned concepts, both about the discount rates and the policies that
can lower their values, reflect onto PRIMES model design. The NEMS model in the US
DOE/EIA follows a quite similar approach, as recommended by Sanstad and
McMahon26.

V.3.c. Discount factors used to evaluate tariffs of using infrastructure

regulated as a natural monopoly.
The model27 uses discount rates based on surveys of actually applied regulated rates
of return by state and regulatory agencies in various countries and for different types
of infrastructure. The surveys indicate that the regulated rates of return on assets of
natural monopolies are set significantly above social discount rates and use the WACC
method. The main difference from private practices is that the state agencies or
regulators do not accept high-risk premium factors on equity capital, in contrast to
private practices. This is justified on the basis that the natural monopoly business has
by definition lower risks compared to business subject to competition.

V.3.d. Business sectors

To determine discount rate values reflecting reality one has to start from a risk-free
(or low risk) discount rate. Business surveys indicate that equity risk premium (which
adds on top of risk free discount rate) is significantly above risk-free rates. The
capitalization structure consisting of borrowed funds at lending interest rate and
equity capital valued at equity risk premium. Practice suggests that companies add
risk premium factors that are specific to a country and/or specific to a sector and/or
specific to a project. Country-specific risk premium are relevant for the short-term

26 “Aspects of Consumers’ and Firms’ Energy Decision-Making: A Review and Recommendations for the
National Energy Modelling System (NEMS)”, Lawrence Berkeley National Laboratory, April, 2008
27 The tariffs of using infrastructure are calculated using the following formula:
𝐶𝑡
𝑅𝐴𝐵 + ∑𝑇𝑡=1
(1 + 𝑑 ± 𝑟)𝑡
𝑃=
𝐷𝑡
∑𝑇𝑡=1
(1 + 𝑑 ± 𝑟)𝑡
RAB is the regulated asset basis (roughly the cumulative cost of investment). 𝐶𝑡 are the annual operating
variable and fixed costs. 𝐷𝑡 denotes the expected future use of the infrastructure (measured as a volume
indicator). 𝑇 is the time horizon. 𝑑 is the regulated discount rate expressing the allowed rate of return on
capital and 𝑟 expresses either a discount on return on capital (if it is deduced) targeted by the regulator or
a bonus (when it is added) used as an incentive for technology or coverage improvement.
and not for long-term projections. Technology-specific risk premium depend on the
degree of technology or market maturity. Sector-specific risk premium differentiates
depending on degree of competition. Therefore, it is lowest for investment by
regulated natural monopolies (e.g., for grids and other infrastructure).

V.3.e. Households
The discount rates used for investment decisions by households in version 6 of the
PRIMES model differentiate by household category based on distributions of
household types. The aim is to capture heterogeneity of behaviours for individual
investment choices by lower or higher income households as well as differentiation
according to type of equipment. Based on the literature, the private discount rates
used by households strongly differentiate by income class. Statistical observations
have confirmed that the discount rates differ by type of energy investment. Surveys
have found lower implicit discount rate values for choice of cars than for housing
equipment. Surveys have also identified that for heating systems and for thermal
integrity expenditures specifically for new-built houses (i.e. choices undertaken when
building the house), the individual discount rates are lower than in similar choices
when renovating existing houses. The reason is that it is more uncertain to undertake
refurbishment investment than incorporating efficient technologies in new houses
taking also into account that the efficiency choices for new houses will last longer than
for existing houses.
For this reason, the model applies lower discount rates (than the default values by
income class) for new buildings concerning thermal integrity and heating systems.
Targeted policies may reduce individual discount rates and such policies can be part
of scenario designs. Policies such as the energy labelling and certain measures
included in Energy Efficiency Directive and the promotion of energy service
companies are examples of such policies.
Version 6 of PRIMES takes into consideration the above-mentioned differentiations of
discount rates by sector. Optionally the model user can modify the discount values by
scenario.

V.4. Use of discount factors for energy system costs reporting

V.4.a. Overview
Once having ran the model for a scenario, which means after simulating behaviours
and market clearing, which are using the discount rates shown in the previous
section, the PRIMES model calculates total energy system costs for reporting
purposes.
In an energy system, there are demanders and suppliers of energy. To assess the cost
impacts from a macroeconomic perspective, the crucial element is the amount that
end use sectors (households and firms, in services and industry, transport and
agriculture) are required to pay in order to get the energy services they need. Energy
services reflect the purpose of using energy, for example, for supporting heating,
cooling, entertainment, mobility and transportation, industrial production, i.e., uses
that enable utility and activity for final energy consumers. Energy services delivery is
due to energy commodities purchased by end-consumers and self-production of
energy. Both depend on energy efficiency at the consumption level.
The end-users undergo investment for purchasing equipment (e.g. boilers, vehicles,
etc.), for insulating buildings and for installing energy saving systems.
From an accounting perspective, the investment expenditures of end-users of energy
are capital expenditures (CAPEX). Part of investment expenditure for equipment
purchasing correspond to energy purposes. For example, the additional cost of a
highly efficient vehicle (on top of cost of a conventional vehicle) incurs for energy
purposes. Only such additional investment costs enter the accounts of energy-related
investment of end-users.
In addition, the final energy consumers incur annual variable and fixed costs which
include the purchasing of energy commodities from energy supplying and trading
sectors, the maintenance costs of equipment and other annual costs (e.g. assurance
costs, vehicle taxes, etc.). These annual costs are operating expenditures (OPEX).
Energy supply and trading sectors fully recover their total costs (CAPEX and OPEX)
from revenues paid by end-consumers. Therefore the total energy system cost only
includes the CAPEX and OPEX incurred by end-consumers, with their OPEX already
incorporating the CAPEX and OPEX costs incurred by the supply and trading sectors.
The PRIMES model determines the prices of supply and trading sectors in a manner
that fully recovers total supply costs using the WACC that represents the real unit cost
of capital experienced by a firm operating in energy supply sectors.
The PRIMES report aggregates CAPEX and OPEX of end-consumers to show a single
total cost figure with annual periodicity. To do this, also the CAPEX figures related to
investments by final energy demand consumers need to be annual following the
equivalent annuity cost method, which involves use of a discount factor over the
lifespan of the investment. The annualised equivalent cost expresses the cost incurred
for the end-consumer for owning an asset until the end of its lifetime. As such it
expresses the gradual accumulation of resources to be able to replace the asset as the
present value of the annuity payments for capital is by definition equal to the
investment (upfront) expenditure (see formulas of equivalent annuity cost method in
Annex I).
The choice of discount rate for the CAPEX cost reporting by final energy demand
consumers can reflect different perspectives, but should reflect in any case the
perspective of the private investor faced with real world investment constraints
In the past, the PRIMES model has used for this cost reporting the opportunity costs of
raising funds as perceived by the end-consumers when making the investment
choices, using the default discount rates by end-consumer for investment decisions in
all scenarios even if in a scenario policy assumptions led to reduced discount rates for
the investment decision. The reason of this choice was to maintain comparability of
total costs across scenarios. This approach has the drawback that perceived high
discount rates is the result of market failures (such as lack of information, split
incentives) which are accounted for as a cost.
 An alternative approach could be to base the cost reporting of the CAPEX by final
energy demand consumers on true payments for capital costs. This implies that the
CAPEX has to be annualised using lending rates for the part of capital borrowed from
banks and equity rates for the rest. It has the drawback that it does not reflect the fact
that there are also opportunity costs associated with higher debt rates (i.e. risk
averseness as well as reduced incentives to make other investments). In addition,
detailed information would need to be collected to identify the borrowing rates faced
by different end-users. Furthermore, equity rates are subjective and therefore
assumptions are necessary regarding their values. Finally, a dilemma similar to that of
the approach using discount rates that take into account opportunity costs arises.
Policies may enable reduction of equity discount rates and if this differs by scenario,
comparability of costs is lost across scenarios.

Summary of cost concepts used to calculate total energy system costs

Final energy consumers Energy supply sectors Total energy system costs

CAPEX Investment expenditures for Investment expenditures for CAPEX incurred directly for final
purchasing equipment, vehicles power generation plants, power energy consumers
and appliances and for thermal grids, gas networks, refineries,
integrity and other energy saving primary fuel extraction, etc.
purposes in the premises of the
consumers

OPEX Purchasing of fuels, distributed Purchasing of fuels and annual OPEX incurred directly for final
heat and electricity, as well as operating and maintenance energy consumers
other annual expenditures for expenditures
operation and maintenance

Profits or Not applicable Applicable to energy supply included indirectly in costs for
deficits of sectors ad network operators purchasing energy commodities by
financial depending on scenario end consumers
balance assumptions about market
distortions

Taxes, Applicable for both CAPEX and Applicable for both CAPEX and Energy tax revenues included.
subsidies and OPEX OPEX Revenues from auctioning ETS
auction allowances not included, assuming
revenues perfect recycling in the economy.

Note: Total CAPEX for the entire energy system is the sum of CAPEX incurred for end-consumers and CAPEX incurred for
Inpublic
energy suppliers, conclusion,
transportcomparability
providers, networkacross the etc.
operators, scenarios is of key importance and implies
that the discount rates used in the cost accounting must not vary between scenarios.
Considering the draw-backs of both approaches listed above it is proposed to account
the costs associated with CAPEX for final energy demand consumers using a lower
rate that is more in line with the WACC used for the supply and industry sector. This
would mean that high perceived discount rates, which may be the result of market
failures (such as lack of information, split incentives), would no longer be accounted
for as a cost, and from a cost accounting perspective would treat demand side sector
and supply side sectors in a similar manner.

V.4.b. Present Value Calculation Method

Assume that there are two options to compare, namely 𝑖 and 𝑗 which require upfront
costs 𝐼𝑖 , 𝐼𝑗 (assumed with a negative sign) and imply variable-operating costs and/or
revenues 𝑣𝑖,𝑡 , 𝑣𝑗,𝑡 where 𝑡 denotes time, which are negative when they are net costs
and positive when they are net revenues.
Net Present Value (NPV) is the difference between the present value of cash inflows
and cash outflows, where future values are discounted using a discount factor 𝑑:
𝑣ℎ,𝑡
𝑁𝑃𝑉ℎ = ∑ + 𝐼ℎ 𝑤𝑖𝑡ℎ ℎ = 𝑖, 𝑗
(1 + 𝑑)𝑡
𝑡
The choice uses the rule that an option is preferable if NPV is positive, otherwise it is
rejected. In case of alternatives, the option with the highest NPV is preferable. In case
all cash flows are negative, (costs) then the option with less negative NPV is
preferable.
The payback period method determines how long it will take to pay back the initial
investment using the sum of expected future revenues. Future revenues may be
savings on future costs that would otherwise occur. This method does not calculate
present values and thus does not use a discount factor. The numerator uses
investment expenditures expressed as a positive amount. The denominator can
include positive and/or negative cash flows (net revenues/savings or net costs,
respectively):
|𝐼ℎ |
𝑃𝐵𝑃 = 𝑤𝑖𝑡ℎ ℎ = 𝑖, 𝑗
∑𝑡 𝑣ℎ,𝑡

If PBP is negative, the project is rejected. To accept a project, it is necessary that it lead
to a PBP, which stands below a threshold that the decision-maker defines as the
maximum reasonably acceptable payback period. The choice of a threshold value is of
the same nature as the choice of the value of a (subjective) discount factor and is also
different by type of actor or project, it involves risk and is conceived relatively to low-
risk (or free-risk) alternatives. If more than one option achieve PBPs below the
threshold value, the option with the lowest PBP (shorter payback period) is
preferable.
The Internal Rate of Return (IRR) method, which in principle is similar to the NPV
method, calculates which discount factor would make a stream of cash flows to have a
NPV equal to zero. The IRR represents how much of a return an investor can expect to
realize from a particular project. The IRR is calculated implicitly by solving the
following problem:
Calculate 𝑑ℎ so that:
𝑣ℎ,𝑡
𝑁𝑃𝑉ℎ = ∑ + 𝐼ℎ = 0 𝑤𝑖𝑡ℎ ℎ = 𝑖, 𝑗
(1 + 𝑑ℎ )𝑡
𝑡

The calculation implies that different IRRs 𝑑ℎ will be obtained for different options,
such as 𝑖 and 𝑗. The IRR method is not applicable to cases having only negative cash
flows (costs). Deciding whether to pursue the investment requires assuming a
threshold value for IRR to retain any project with IRR above the threshold value and
in case of multiple options retain the one with highest IRR above the threshold value.
The choice of the threshold value is subjective, involves risk consideration and has a
relative meaning with reference to low risk or free-risk options.
The equivalent annuity cost method expresses the NPV as an annualized cash flow
by dividing it by the present value of the annuity factor. The annuity factor involves
assumption of a discount factor (𝑑), which as before is subjective, involves risk
considerations and is relative to low or free-risk options. This method is appropriate
when assessing only the costs of specific investment options that have constant or
slightly changing annual variable and operating costs over time. In this form, it is
known as the equivalent annuity cost (EAC) method and represents the cost per year
of owning and operating an asset over its entire lifespan. For example, if the annual
variable operating cost 𝑣ℎ,𝑡 is the same over time and 𝐿ℎ is the economic lifetime of
the project ℎ, then the annual total cost of the asset is:
𝑑
𝐸𝐴𝐶ℎ = 𝐼ℎ ∙ + 𝑣ℎ,𝑡 𝑤𝑖𝑡ℎ ℎ = 𝑖, 𝑗
1 − (1 + 𝑑)−𝐿ℎ
In this formula, 𝐼ℎ and 𝑣ℎ,𝑡 are assumed to be positive and to represent costs. The EAC
method applies when asset selection performs from a cost minimisation perspective.
The EAC method does not say whether it is worth investing, as this requires
consideration of revenues from the asset. It is easy to verify that the present value of
the equivalent annuity costs are exactly equal to initial investment.
Although the above-mentioned capital-budgeting methods seem different to each
other, in essence, they are quite similar and all involve, explicitly or implicitly,
consideration of a discount factor or a threshold value, which are subjective, involve
risk and are relative to risk-free options.

V.4.c. Mathematical illustration of cost reporting in PRIMES

Assume that 𝑡 denotes time and 𝑡𝑡 represents the time of implementing a new
investment. Obviously 𝑡𝑡 ≤ 𝑡. Based on model projections for a scenario, the cost
reporting routine knows 𝐶𝐴𝑃𝐸𝑋𝑖,𝑡𝑡 representing investment expenditure
implemented in 𝑡𝑡 by an end-consumer, with 𝑖 spanning all such investment cases.
The end-consumers also incur variable and fixed annual operating and maintenance
costs represented by 𝑂𝑃𝐸𝑋𝑖,𝑡 . The CAPEX are annualised as 𝐴𝑛𝑛𝐶𝐴𝑃𝐸𝑋𝑖,𝑡 following
the equivalent annuity cost method, using a discount factor 𝑑𝑖 specific to 𝑖 with
lifetime 𝐿𝑖 , as follows:
𝑑𝑖
𝐴𝑛𝑛𝐶𝐴𝑃𝐸𝑋𝑖,𝑡 = ∑ 𝐶𝐴𝑃𝐸𝑋𝑖,𝑡𝑡 ∙ ∙ (1 if 𝑡𝑡 + 𝐿𝑖 ≤ 𝑡, or 0 otherwise)
1 − (1 + 𝑑𝑖 )−𝐿𝑖
𝑡𝑡≤𝑡

Total annual energy system costs 𝐴𝑛𝑛𝑇𝐶𝑡 is:

𝐴𝑛𝑛𝑇𝐶𝑡 = ∑(𝐴𝑛𝑛𝐶𝐴𝑃𝐸𝑋𝑖,𝑡 + 𝑂𝑃𝐸𝑋𝑖,𝑡 )

𝑖

Total cumulative energy system cost 𝐶𝑢𝑚𝑇𝐶 derives as present value of total annual
energy system costs using a social discount rate equal to 𝛿:

𝐶𝑢𝑚𝑇𝐶 = ∑ 𝐴𝑛𝑛𝑇𝐶𝑡 ∙ (1 + 𝛿)−𝑡

𝑡

Notice that 𝛿 is different from the private discount rates 𝑑𝑖 used to annualise CAPEX.
Therefore 𝐶𝑢𝑚𝑇𝐶 evaluated for different scenarios ranks the scenarios from a social,
public policy perspective.