Energy 239 (2022) 122253

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Is total system cost minimization fair to all the actors of an energy

system? Not according to game theory
Robert Fischer*, Andrea Toffolo
Energy Engineering, Division of Energy Science, Luleå University of Technology, SE-97187, Luleå, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: A common approach to energy system optimization is to minimize overall costs at system level,
Received 16 February 2021 regardless of the actors actually bearing those costs. This paper presents an approach inspired by Nash
Received in revised form game theory concepts, in which the actors involved in an energy system determine their optimal stra-
1 September 2021
tegies according to their own economic interests (profit functions) in a non-cooperative or in a coop-
Accepted 3 October 2021
erative way. A simple case study, considering an electric utility and individual heating consumers in the
Available online 5 October 2021
municipal energy system of a small town in northern Sweden, shows the differences between the two
approaches. The game theory approach is able to represent more realistic interactions among the actors
Keywords:
Energy system optimization
of an energy system, fair in fulfilling their conflicting economic interests, and, therefore, a more suitable
Game theory tool for decision makers evaluating the impacts of policy instruments.
Profit functions © 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license
Nash equilibrium (http://creativecommons.org/licenses/by/4.0/).
Nash bargaining solution
Policy instruments

1. Introduction variables (e.g., the design parameters of a thermodynamic cycle, the

location and the size of a new biorefinery, how much electricity will
Energy system optimization (ESO) is a fundamental tool to be generated in a region or a country using a particular energy
support researchers and decision-makers in shaping a sustainable conversion technology). Given the environmental targets for the
and decarbonized future. Achieving ambitious environmental tar- system, these actors try to act according to their own best economic
gets within a compelling timeline requires the evaluation of future interest, although the achievement of environmental targets goes
scenarios and/or transition pathways with digital models and the in general against that.
interpretation of the results obtained from the optimization prob- Clearly, minimizing a single economic objective function cannot
lems formulated through those models. represent fairly the economic interests of all the actors at the same
ESO is usually performed on mixed integer linear models of time. At best, the optimal solution found will coincide with the
systems encompassing, e.g., industrial sites integrating different interest of one or few of them. The actors being dissatisfied by this
plants [1], supply chains [2,3], or local, regional, national [4] and optimal solution will tend to behave differently in the real world. If
even international energy systems. A single economic objective they actually choose a different strategy, the system will evolve
function is minimized subject to several constraints, among which towards a different solution than the one predicted by using a “total
those related to environmental targets. This objective function is system cost” minimization approach.
named “total system cost” or “societal cost” and is the sum of all the The remedy proposed in this paper is to embrace the different
investment and operating costs related to processes/technologies/ points of view of the actors involved in the system, abandoning the
sectors within system boundaries. artifice of a “total system cost”, and to apply game theory concepts
However, in any modeled energy system it is possible to identify in order to attempt to a fair harmonization among these points of
a number of actors involved, those which in the real world will take view, i.e. to achieve a fair allocation of the increased costs due to the
the decisions about the strategies setting the values of some model transition to sustainable and decarbonized energy systems among
the actors.
The open literature shows that the application of game theory
concepts to ESO has been mainly focused since the early 2000s on a
* Corresponding author. particular niche, power system management, for investigating the
E-mail address: robert.fi[email protected] (R. Fischer).

https://doi.org/10.1016/j.energy.2021.122253
0360-5442/© 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
R. Fischer and A. Toffolo Energy 239 (2022) 122253

equilibrium resulting from the operation strategies of direct com- Game theory aims at identifying the rational decisions made in a
petitors (hence a set of actors with very similar roles in the system). given context (game) by agents (players) that are interdependent,
For instance, electricity price is determined in Ref. [5] as the result i.e., they influence one another's outcome (gain) through their
of the competition among companies in deregulated markets and decisions (strategies).
the market clearing price is found in Ref. [6] as the result of the In general, a game is defined by:
interaction among smart microgrids. The operators of power plants
with different technologies in an integrated energy system are the - A list of players Pi, with i ¼ 1,2, …,n, each aiming at maximizing
competing actors on the supply side in Ref. [7], and a similar their own gain;
application can be found in Ref. [8] about the dispatch of heat in a - The sets of strategies available to each player S1,S2, …Sn (si 2 Si is
distributed energy network where different buildings (a store, a a strategy that player Pi can choose);
hospital, a hotel) are the actors. - The gains of each player Gi (s1,s2, …,sn), i.e. mathematical ex-
Very little has been presented about more general energy sys- pressions defined as functions of the strategies chosen by the
tems, in particular related to their net zero carbon transition. Even players.
when the energy system model is coupled to a partial equilibrium
market model to evaluate the effects on the prices of resources and A generic solution of a game is represented by a combination of
intermediate and final products, the minimization of supply chain the strategies chosen by the players s¼(s1,s2, …,sn), also called
total cost is still used to drive the top-level optimization procedure “strategy profile”, and the corresponding values of the gains Gi(s).
[9]. PRIMES (Price-Induced Market Equilibrium System [10]) is the Games can be non-cooperative or cooperative, leading to solu-
only framework for ESO in which actors representing different tions having very different features.
sectors of a national or international energy system (e.g. house-
holds, steel industry, trade sector, power generation, etc.) maximize 2.1. Non-cooperative games e The (generalized) Nash equilibrium
individually their own profit or welfare. PRIMES electricity and gas
market modules apply Nash-Cournot competition and a Nash In non-cooperative games the players have to make their own
equilibrium is found when the model is solved. The PRIMES decisions without knowing the strategies chosen by the other
framework was used to obtain the trend projections on transport, players (although in “complete information” games each player has
energy and CO2 emission in the EU Reference Scenario 2016 [11,12]. perfect knowledge of the strategy sets and gain functions of the
However, the available documentation (PRIMES is not public other players).
domain) does not clarify whether other types of interactions (e.g. In this situation the concept of Nash equilibrium [14] describes a
cooperation or coalitions) among the actors are considered. game solution in which no player can increase their gain by
This paper aims to: modifying their chosen strategy given the strategies chosen by the
other players. Formally, a strategy profile (s1*,s2*, …,sn*) is a Nash
- compare the solution obtained by minimizing “total system equilibrium if, for each player Pi, Gi (s1*, …,si*, …,sn*)  Gi (s1*, …,si,
cost” with those representing the best economic interests of the …,sn*) for each si 2 Si.
actors involved in the system (which may have different roles, In games with coupling constraints, where the strategies avail-
e.g., suppliers vs. users, public vs. private investors); able to a player may be limited by the strategies chosen by the other
- apply game theory concepts to ESO in order to identify a fair players, a strategy profile (s1*,s2*, …,sn*) is a generalized Nash
harmonization among the economic interests of the actors in equilibrium if, for each player Pi, Gi (s1*, …,si*, …,sn*)  Gi (s1*, …,si,
the transition towards sustainable and decarbonized energy …,sn*) for each si belonging to the subset of strategies available to
systems; player Pi given (s1*, …,si-1*,siþ1*, …,sn*).
- discuss key differences between the traditional approach using It is maybe worth to remember here that there are games in
a “total system cost” function and the game theory approach which a (generalized) Nash equilibrium does not exist, or is not
using profit functions expressing the economic interests of the unique, and that a (generalized) Nash equilibrium is not necessarily
actors. Pareto optimal.

The case study considered in this paper to compare the tradi- 2.2. Cooperative games e The Nash bargaining solution
tional total system cost minimization approach and the game
theory approach is the municipal energy system of a town in In cooperative games the players agree to coordinate their
northern Sweden (Piteå), comprising the electricity generation strategies in order to achieve higher gains than those obtainable
sector and the individual heating sector. Two actors (the municipal without an agreement. If the players did not agree to cooperate, the
energy utility and a group of consumers) will have to decide their result would be the disagreement outcome of the game, which is
strategies to achieve a target on the reduction of the CO2 emissions denoted with the strategy profile d¼(d1,d2, …,dn) where di 2 Si.
for which the municipal energy system is responsible. If the players agree to cooperate, the candidate game solutions
The remainder of the paper is structured as follows. Relevant are those obtained from a multi-objective optimization problem in
definitions and concepts from game theory are illustrated in Sec- which the objective functions are the gain functions of the players.
tion 2, while the methodology and the case study are described in The strategy profiles p belonging to the Pareto optimal set POS of
Section 3. Section 4 presents the results about the case study, which such optimization problem are the best tradeoffs among the ob-
are discussed, together with remarks from a more general jectives, so they also represent the most convenient agreements
perspective, in Section 5. Finally, the conclusions are presented in among the players. However, since gains Gi are conflicting objec-
Section 6. tives, the overall benefit deriving from the fact that players agree to
coordinate their strategies is allocated in different ways among the
2. Game theory concepts players when different optimal agreements are considered,
although they are equally optimal in a Pareto sense.
The fundamental concepts of game theory that will be used in The fairest allocation for all players is then identified by the
this paper are briefly introduced in this section. The reader is Nash bargaining solution [15], a strategy profile p* among those of
referred to textbooks on the subject for more details (e.g. Ref. [13]). the Pareto optimal set so that:
2
R. Fischer and A. Toffolo Energy 239 (2022) 122253

Two actors (¼players) are identified in this municipal energy

system, the “Utility” and the “Consumers”.
The “Utility” is a single, vertically integrated municipal energy
utility that invests in local renewable electricity generation, im-
ports/exports electricity from/to the grid and supplies electricity to
industrial and non-industrial customers. The strategy of the Utility
for reducing CO2 emissions is to invest in local electricity genera-
tion from renewable, emission free sources, i.e. renewable energy
(RE) technologies. The strategy of the Utility can be described using
the following decision variables:

- Onshore wind power capacity, from the 145 MW already

installed by 2015 to 865 MW in steps of 3.6 MW (a typical
onshore wind turbine), considering actual project de-
velopments [18];
- Offshore wind power capacity, from 0 to 600 MW in steps of
6 MW (a typical offshore wind turbine), considering offshore
Fig. 1. Nash bargaining solution p* identified in the Pareto front of players' gains for a
given disagreement outcome d. project proposals [19];
- Solar photovoltaic (PV) capacity, from 0 to 1000 MW in steps of
1 MW (modelling large scale solar PV plants), corresponding to a
Y
p* ¼ arg max ðGi ðpÞ  Gi ðdÞÞ (1) maximum land use of 0.3% within Piteå municipality.
p2POS
i¼1;n
The “Consumers” are the customers of the individual heating
According to this definition, the Nash bargaining solution sector, i.e. the households not connected to the district heating. The
maximizes the product of the additional gains each player gets, strategy of the Consumers for reducing CO2 emissions is to lower
thanks to their cooperation, with respect to the disagreement their demand of electricity for heating, choosing among different
outcome. Please note that, contrary to Nash equilibrium, the Nash heating technologies and/or investing in energy efficiency mea-
bargaining solution can be proven to exist and be unique. sures in buildings. The strategy of the Consumers can be described
From a geometrical perspective, in a two player game the using the following decision variables:
agreement corresponding to the Nash bargaining solution is iden-
tified as the point in which an equilateral hyperbola with origin in - The quota of the individual heating sector demand satisfied with
the disagreement outcome (G1(d),G2(d)) is tangent to the Pareto direct electric heating (electric boilers or electric wall panels);
front (see Fig. 1). In fact, an equilateral hyperbola represents the - The quota of the individual heating sector demand satisfied with
locus of all points having a certain value of the product of the heat pumps (a mix of air to air and ground source heat pumps,
additional gains (G1(s)-G1(d)) (G2(s)-(G2(d)), and the equilateral with an average COP equal to 3)
hyperbola that maximizes this product for the Pareto optimal - The quota of the individual heating sector demand satisfied with
strategy profiles must be the one tangent to the Pareto front (the biomass heating (pellet boilers)
value of the product would be lower for any other equilateral hy- - The reduction of the individual heating sector demand due to
perbola closer to the origin, i.e. intersecting the Pareto front). the implementation of energy efficiency measures in buildings.
The relationship between these energy savings and the related
3. Methodology costs has been investigated in Ref. [20] (projects Halvera Mera
and BETSI) [21e25], and is illustrated in Fig. 2.
3.1. Case study
Please note that, due to the size of the actors, the strategies
The case study used in this paper to compare the traditional chosen by the Utility and the Consumers are not going to affect the
total system cost minimization approach and the game theory
approach is the municipal energy system in Piteå, a small town
located in the north of Sweden. It is modeled using the simulation
tool EnergyPLAN [16], which provides energy balances, system
costs and environmental parameters with hourly granularity over a
one-year period. The model comprises and integrates the electricity
and heating sectors within the borders of the municipality. In the
electricity sector, local electricity generation (hydro- and wind
power, industrial bio-CHP) is renewable and emission free. Power
balance is achieved by using the grid for importing the deficit or
exporting the surplus from local electricity generation. The heating
sector is divided into two parts: district heating and individual
heating of buildings. District heating is emission free, supplied with
biomass and with waste heat from the local pulp and paper in-
dustry. Individual heating relies on biomass heating and electric
heating (direct or heat pumps). This model is the context (¼game)
in which the optimization problems considered in this study are
formulated. More details about the municipal energy system in
Piteå and its model in EnergyPLAN can be found in the Supple-
mentary material to this paper and in Ref. [17]. Fig. 2. Investment costs for energy efficiency measures in a single-family home.

3
R. Fischer and A. Toffolo Energy 239 (2022) 122253

market prices of technologies and resources (electricity and and 40 EUR/MWh) and two different levels of subsidies for the
biomass). costs of the energy efficiency measures in buildings (50% and 75%).
An environmental target is set to reduce by 50% the CO2 emis-
sions for which the municipal energy system is responsible, and it is
4. Results
implemented as a constraint in the optimization problems. In this
model of the Piteå municipal energy system, it is clear that the only
4.1. Minimization of total system cost (traditional ESO approach)
CO2 emissions are due to the import of electricity from the grid,
according to a constant grid emission factor calculated for the
The results obtained in the considered four cases by minimizing
Nordic electricity mix (0.156 tCO2eq/MWh) [26]. As a consequence,
the total system cost function defined in EnergyPLAN under the
any variation of supply and/or demand decreasing electricity
constraint of achieving a reduction of at least 50% of CO2 emissions
import (the utility deciding to invest in a higher RE capacity or the
are shown in Table 1.
consumers deciding to decrease their electricity demand for indi-
In the individual heating sector the Consumers are expected to
vidual heating by choosing a different heating technology or by
use heat pumps only, since the switch from electric boilers to heat
investing more in energy efficiency measures) will result in lower
pumps is a low hanging fruit for reducing CO2 emission, while the
CO2 emissions, and vice versa.
switch to biomass boilers would occur only with much more severe
targets on CO2 emissions (well above a 75% reduction, see Ref. [17]).
3.2. Optimization problems for traditional and game theory
Energy efficiency measures in buildings can somewhat reduce the
approaches to ESO
overall individual heating demand (about 212 GWh) and, therefore,
the demand of electricity from the heat pumps, but the Consumers
The traditional approach to ESO consists in the minimization of
will invest in these measures only with a convenient level of sub-
the “total system cost” function, which in this case study is defined
sidies (which depends on the associated discount rate and the ESP).
(for one year) as the sum of:
In this case study, the Consumers would consider investing in en-
ergy efficiency measures at a subsidy level of 75% regardless of the
- Annualized costs of the installed RE technologies (based on
considered ESP (but discount rates lower than 9% would lower this
capital cost, expected lifetime, discount rate 9%, operation and
threshold).
maintenance costs);
In the electricity sector the Utility is expected to invest only in
- Annualized costs of the individual heating devices (based on
onshore wind turbines when the ESP is 40 EUR/MWh, very close to
capital cost, expected lifetime, discount rate 9%, operation and
the levelized cost of electricity for this technology considering a
maintenance costs);
discount rate of 9% (for discount rates lower than 6% the levelized
- Fuel (pellet) costs for biomass individual heating;
cost of electricity of onshore wind power is below 40 EUR/MWh).
- Costs/revenues from the import/export of electricity from/to the
Otherwise, a mix of onshore wind turbines and solar PV should be
grid, both calculated according to the electricity spot price (ESP).
installed to take advantage of the different seasonal profiles of
electricity generation. For both the considered ESP, offshore wind
On the other hand, a game theory approach to ESO focuses on
turbines are too expensive to be installed until very high CO2
the point of view of the actors involved in the system, i.e. on their
emissions reductions are considered (again, well above 75% [17]).
economic interests represented by their own profit functions
(¼gains). For the utility, the profit function (for one year) is defined
as the sum of: 4.2. Maximization of the profit functions

- Revenues from electricity sales to customers (based on the ESP); Fig. 3 shows the Pareto optimal sets and fronts obtained in the
- Revenues from electricity export to the grid (based on the ESP); considered four cases by maximizing the profit functions of the two
- Revenues from grid fees; actors under the constraint of achieving a reduction of at least 50%
- Annualized costs of the installed RE technologies (based on of CO2 emissions.
capital cost, expected lifetime, discount rate 9%, operation and Maximum profit for the Utility is found when the consumers use
maintenance costs); only biomass boilers, i.e. when the electricity demand for individ-
- Costs due to electricity import from the grid (based on the ESP). ual heating is reduced to zero. In fact, both the considered ESP are
below the levelized cost of electricity of the available RE technol-
For the consumers, the profit function (for one year) is actually ogies (with a 9% discount rate), so that installing more RE capacity,
made of costs only (all consumer costs include VAT): as required for complying with the target on CO2 emissions
reduction, always goes against the economic interest of the Utility.
- Annualized costs of the individual heating devices (based on Minimum costs for the Consumers are found when only heat
capital cost, expected lifetime, discount rate 9%, operation and pumps are used for individual heating. This condition also
maintenance costs);
- Annualized costs of energy efficiency measures in buildings
Table 1
(based on capital cost, expected lifetime, discount rate 9%,
Results with total system cost minimization.
operation and maintenance costs);
- Costs for fuel (pellet) for biomass individual heating (including ESP [EUR/MWh] 25 25 40 40

delivery to home); Subsidy for energy efficiency measures 50% 75% 50% 75%
- Costs for electricity for electric individual heating (including Onshore wind power capacity [MW] 280 270 330 319
Offshore wind power capacity [MW] 0 0 0 0
grid fees and energy taxes).
Solar PV capacity [MW] 33 33 0 0
Electric boilers [GWh] 0 0 0 0
In this approach, a two-objective optimization problem is set up Heat pumps [GWh] 212 185 212 190
to maximize the profit functions of the two actors at the same time. Biomass boilers [GWh] 0 0 0 0
The optimization problems related to the two approaches are Savings from energy efficiency measures [GWh] 0 27 0 22
Total system cost [MEUR] 84.6 84.2 81.5 81.9
solved in four cases, assuming two different values of the ESP (25
4
R. Fischer and A. Toffolo Energy 239 (2022) 122253

Fig. 3. Pareto optimal sets and fronts obtained by maximizing utility profit and minimizing consumer costs, complying with the environmental target of reducing local CO2
emissions by at least 50%.

corresponds to the lowest profit for the Utility, because the elec- for individual heating. If the energy efficiency measures in build-
tricity demand (and therefore the need to install a higher RE ca- ings get a 75% subsidy, then an equilibrium can be found between
pacity) would increase only if the Consumers used a mix of electric the savings from electricity and the investments in the less
boilers and heat pumps, but the Consumers do not find electric expensive types of renovations, reducing the annual heating de-
boilers more convenient than heat pumps due to the high operating mand from 212 GWh to about 179 GWh. On the other hand, the
costs. Utility should install only onshore wind turbines when the ESP is 40
Between these two extremes a mix of heat pumps (less EUR/MWh, whereas a mix of onshore wind turbines and solar PV
expensive) and biomass boilers (more expensive) should be used should be installed when the ESP is 25 EUR/MWh. The installed

5
R. Fischer and A. Toffolo Energy 239 (2022) 122253

capacity has to grow with the electricity demand from the indi- a disagreement outcome (DO) that will serve as a reference strategy
vidual heating sector in order to keep the electricity import con- profile for determining the Nash bargaining solution. This can be
stant and comply with the target on CO2 emissions reduction. done by analyzing the options of the two actors, in light of the
following considerations:
4.3. Non-cooperative and cooperative strategies of the actors
- achieving the target on CO2 emissions reduction takes prece-
If a non-cooperative game is considered, a quite uncommon dence over each actor trying to maximize its own profit;
situation occurs in this particular case study. All the points in the - CO2 emission reduction goes against the economic interest of
Pareto front correspond to strategy profiles that are generalized each actor;
Nash equilibria, given that the Pareto front is formed because of - the resulting level of CO2 emissions reduction depends on the
(and therefore along the boundary of) the constraint on CO2 interaction of the strategies chosen by the actors.
emissions reduction.
In fact, starting from a given point of the Pareto front, the The Consumers alone are not able to reach the target on CO2
Consumers could reduce their costs only by using more heat pumps emissions reduction, because the impact of their decisions on the
instead of biomass boilers (but this would increase the electricity amount of imported electricity is limited. If they decide to use
demand with a fixed supply by the Utility, increasing electricity biomass boilers only (with or without renovations) while the Utility
import and therefore violating the constraint on CO2 emissions does not install new renewable energy capacity, CO2 emissions
reduction) and the Utility could increase the profit by reducing the reduction would be insufficient (only 22%). Accordingly, their best
installation of new renewable energy capacity (but this would decision (not knowing the strategy of the Utility) is to use biomass
reduce the supply with a fixed demand, increasing electricity boilers only, with some renovations if these are economically
import and therefore violating the constraint, too). convenient. In this way, they contribute as much as they can to CO2
In this particular case study, considering a non-cooperative emissions reduction, even though their costs will be higher,
game is not helpful, because the two actors, seeking separately an whereas higher investments in renovations would only increase
equilibrium between their strategies, may end up in any of the these costs.
strategy profiles of the Pareto optimal set. Unfortunately, this does On the other hand, the Utility (not knowing the strategy of the
not provide any information on the fairness of equilibrium reached Consumers) must assume that they might decide to just consider
(if the equilibrium were unique, it would necessarily be fair). their own profit and use heat pumps only, with some renovations if
Let us consider then a cooperative game, establishing first of all these are economically convenient. Accordingly, the best decision

Fig. 4. Utility profit vs consumer costs: disagreement outcome (DO) and Nash bargaining solution (NBS).

6
R. Fischer and A. Toffolo Energy 239 (2022) 122253

for the Utility is to install exactly the renewable energy capacity reducing the overall electricity demand (which in turn would
that is needed to comply with the target on CO2 emissions reduc- result in a lower electricity import and lower CO2 emissions
tion in case the Consumers acted in their own best economic in- from the grid mix). On the contrary, biomass boilers are the
terest. In this way, the Utility is sure that the target will be achieved more expensive heating technology, so that the reduction in the
regardless of the strategy chosen by the Consumers, even though consumer demand for individual heating electricity is just
the profit for the Utility will be negatively affected, whereas the driven by the environmental target and is not economically
installation of a larger amount of renewable energy capacity would convenient for the Consumers.
further decrease this profit. - If selling the generated electricity surplus could result in a profit,
As illustrated in Fig. 4, from the values of the profit functions or if the spot price were higher than the LCOE of RE technolo-
calculated for the disagreement outcome it is possible to determine gies, then the Utility would naturally install as much capacity as
the Nash bargaining solution. In all the four considered cases the possible (which also would result in a lower electricity import
Nash bargaining solution sees a large share of the Consumers using and lower CO2 emissions from the grid mix). On the contrary,
biomass boilers instead of heat pumps. This share (see Fig. 5) can be generation surplus is not profitable and the ESP is lower than
estimated as roughly around one fourth for the higher value of the the LCOE of RE technologies, so that the Utility generates only as
ESP and around two fifths for the lower one, whereas it is not much electricity as needed to reach the target on CO2 emissions
significantly affected by the subsidies to energy efficiency measures reductions (because a higher production would not generate
in buildings. more profit).

The comparison in Fig. 5 makes it clear that for all the four
5. Discussions
considered cases the solutions minimizing total system cost at 50%
CO2 emissions reduction are different from the Nash bargaining
Before going into specific comments about the results of this
solutions obtained considering the profit functions of the Utility
case study and then enlarging the perspective towards a more
and the Consumers (and calculated according to their disagreement
general discussion, it is perhaps worth to review the conditions that
outcomes). The decision variables of the solutions minimizing total
make this case study significant.
system cost at 50% CO2 emissions reduction are actually close to
The environmental constraint (in this case the target on CO2
those minimizing the costs for Consumers, meaning that the total
emissions reduction) forces the two actors to find an equilibrium or
system cost function negatively affects the Utility in favor of the
a fair harmonization between their strategies only if reaching the
Consumers. This proves the limited ability of a single economic
target goes against the best economic interest of both actors:
objective function to capture the complexity of the interactions
among the actors even in simple energy systems (as the one
- If biomass boilers were cheaper than any electric heating option,
considered in this paper).
the Consumers would naturally use 100% biomass boilers

Fig. 5. Comparison between the optimal solutions minimizing total system cost at 50% CO2 emission reduction and the Nash bargaining solutions.

7
R. Fischer and A. Toffolo Energy 239 (2022) 122253

In fact, a total system cost function is structurally different from their strategies (a non-cooperative game, leading to Nash equilib-
the profit functions of the actors, as the latter include terms that are rium/a) and the other in which the actors mutually coordinate their
not considered in the former. In this case study, for instance, the strategies (a cooperative game, leading to a particular solution
VAT paid by the Consumers and, even more significant, the transfer among the trade-offs among the best economic interests of the
of money between the Utility charging the electric bills and the actors, the Nash bargaining solution).
Consumers paying for them alter the relative importance that other The results of the rather simple case study presented in this
terms have in the total system cost function, such as the investment paper showed that the optimal solutions obtained by minimizing a
costs or the cost for the biomass fuel. This is clearly the main cause total system cost function are different from those obtained by
of the differences shown in Fig. 5. The role and choice of the dis- maximizing the profit functions of the actors involved in the sys-
count rate in a total system cost function, usually a “social” discount tem. The most significant difference is that the consumers of the
rate uniformly applied to all investments (as in EnergyPLAN), is individual heating sector in Piteå municipality should use only heat
debatable too [27e32]. In a private company, the discount rate pumps according to total system cost minimization, but according
must be equal or higher than the weighted average cost of capital to the game theory approach a consistent share of them, from 25%
for the investment to be approved [11,33]. For wealthy households to 40% depending on the considered ESP, would use biomass boilers
the implicit discount rate can be close to zero for energy efficiency instead. The cause of this difference was identified in the modified
investments, whereas for low-income households it can be up to relative importance of some of the terms that appear in the total
100% and higher [34,35]. With reference to the presented case system cost function.
study, the energy efficiency measures in buildings must be heavily It can be concluded that a game theory approach to ESO helps to
subsidized to reduce the demand in the individual heating sector identify the factors that actually determine the optimal strategies of
(at a 9% discount rate and for both the considered ESPs, a household the actors, considering their interactions in a more realistic way,
would only invest in energy efficiency measures that are 75% and leads to a fair harmonization of the economic interests of the
subsidized). The required subsidy level would drop if the discount actors. Among these factors, those introduced by policy in-
rate for the consumers were lower (e.g., to 50% at a discount rate of struments are of special interest to the decision makers planning
4%, see also [36] on the importance of discount rate in evaluating energy transition scenarios.
the economic viability of energy efficiency measures in buildings). In future works, different policy instruments will be assessed
A final remark on the differences between a total system cost with this game theory approach, investigating their intended and
minimization approach and a game theory approach to ESO is unintended outcomes in order to contribute to the knowledge
about their radically different nature. Thanks to the analysis of the development on an optimal mix of policies that is suitable to ach-
strategies available to the actors, a game theory approach to ESO is ieve the transition to sustainable and decarbonized energy systems
not only able to represent more realistic interactions among the with balanced socio-economic solutions. The policy instruments to
actors of an energy system, but it offers to decision makers a larger be assessed are mainly those affecting the economic interests of the
spectrum of information, which is not limited to the Pareto optimal actors, such as grants, subsidies, feed-in tariffs, carbon taxes or tax
solutions obtained from the simultaneous maximization of the reliefs, but also mandates including obligation for solar water
profit functions of the actors. Valuable knowledge is provided by heating or the ban of fossil fuel heating.
evaluating the alternative between the non-cooperative option,
resulting in the Nash equilibrium (in spite of the limited usefulness Author contribution
in the presented case study), and the cooperative option, resulting
in a fair allocation of the associated mutual benefit in the Nash Fischer R: Conceptualization; Data curation; Formal analysis;
bargaining solution, and this is the basis for non-trivial decisions Investigation; Methodology; Validation; Software; Writing - orig-
that must be taken by the actors. From this perspective, the mini- inal draft; Toffolo A: Conceptualization; Formal analysis; Method-
mization (or maximization) of a single system-level economic ology; Software; Supervision; Writing - review & editing.
objective function, even if it could represent the sum of the costs (or
profits) of all the actors involved, may be seen as the external Declaration of competing interest
imposition of a sort of forced cooperation among the actors for a
greater good, a cooperation in which the actual effects on the single The authors declare that they have no known competing
actors are overlooked. financial interests or personal relationships that could have
A corollary to this remark is that a game theory approach to ESO appeared to influence the work reported in this paper.
is a more suitable to assess the impact of policy instruments,
because it focuses on the strategies of the different actors as they Appendix A. Supplementary data
are moved by their economic interests. Policy instruments are
typically designed to affect specific actors of the energy system Supplementary data to this article can be found online at
(e.g., feed-in tariffs for utilities, subsidies or tax reliefs for house- https://doi.org/10.1016/j.energy.2021.122253.
hold energy efficiency measures), but may also have unintended
consequences for other actors. A game theory approach is capable References
of providing indications on how to avoid unintended and/or unfair
allocations of additional costs or benefits, and allows to analyze [1] Kantor I, Robineau JL, Bütün H, Mare
chal F. A mixed-integer linear program-
ming formulation for optimizing multi-scale material and energy integration.
more specifically the effectiveness of financial policy measures. Front Energy Res 2020;8. https://doi.org/10.3389/fenrg.2020.00049.
[2] Leduc S. Development of an optimization model for the location of biofuel
6. Conclusions production plants. PhD thesis. Luleå University of Technology; 2009.
[3] Pettersson K, Wetterlund E, Athanassiadis D, Lundmark R, Ehn C, Lundgren J,
et al. Integration of next-generation biofuel production in the Swedish forest
This paper showed how game theory concepts can be applied to industry - a geographically explicit approach. Appl Energy 2015;154:317e32.
ESO, considering the best economic interests of the actors involved https://doi.org/10.1016/j.apenergy.2015.04.041.
in the system instead of the optimization of a single economic cost [4] Krook-Riekkola A. National energy system modelling for supporting energy
and climate policy decision-making: the case of Sweden. PhD thesis. Chalmers
function, usually expressing overall costs at system level. Two University of Technology; 2015.
contexts were considered, one in which the actors do not agree on [5] Sarjiya, Budi RFS, Hadi SP. Game theory for multi-objective and multi-period

8
R. Fischer and A. Toffolo Energy 239 (2022) 122253

framework generation expansion planning in deregulated markets. Energy [21] Hirvonen J, Jokisalo J, Heljo J, Kosonen R. Towards the EU emission targets of
2019;174:323e30. https://doi.org/10.1016/j.energy.2019.02.105. 2050: cost-effective emission reduction in Finnish detached houses. Energies
[6] Hasankhani A, Hakimi SM. Stochastic energy management of smart microgrid 2019;12. https://doi.org/10.3390/en12224395.
with intermittent renewable energy resources in electricity market. Energy [22] Hirvonen J, Jokisalo J, Heljo J, Kosonen R. Towards the EU emissions targets of
2021;219. https://doi.org/10.1016/j.energy.2020.119668. 2050: optimal energy renovation measures of Finnish apartment buildings.
[7] Wang H, Zhang C, Li K, Ma X. Game theory-based multi-agent capacity opti- Int J Sustain Energy 2019;38:649e72. https://doi.org/10.1080/
mization for integrated energy systems with compressed air energy storage. 14786451.2018.1559164.
Energy 2021;221. https://doi.org/10.1016/j.energy.2021.119777. [23] Discher H, Hinz E, Enseling A. dena-Sanierungsstudie. Teil 1: wirt-
[8] Wu Q, Ren H, Gao W, Ren J. Benefit allocation for distributed energy network schaftlichkeit energetischerModernisierung im Mietwohnungsbestand. 2010.
participants applying game theory based solutions. Energy 2017;119:384e91. Berlin.
https://doi.org/10.1016/j.energy.2016.12.088. [24] Stolte C, Marcinek H, Discher H, Hinz E, Enseling A. dena-Sanierungsstudie.
[9] Zetterholm J, Bryngemark E, Ahlstro €m J, So
€ derholm P, Harvey S, Wetterlund E. Teil 2: wirtschaftlichkeit energetischer Modernisierung in selbstgenutzten
Economic evaluation of large-scale biorefinery deployment: a framework Wohngeb€ auden. 2011. Berlin.
integrating dynamic biomass market and techno-economic models. Sustain [25] Shadram F, Bhattacharjee S, Lidelo € w S, Mukkavaara J, Olofsson T. Exploring
Times 2020;12. https://doi.org/10.3390/su12177126. the trade-off in life cycle energy of building retrofit through optimization.
[10] E3M-Lab: National technical university of athens department of electrical and Appl Energy 2020;269. https://doi.org/10.1016/j.apenergy.2020.115083.
computer engineering. The PRIMES Model n.d. http://www.e3mlab.eu/ [26] Moro A, Lonza L. Electricity carbon intensity in European Member States:
e3mlab/(accessed February 12, 2021). impacts on GHG emissions of electric vehicles. Transport Res Transport En-
[11] Zampara M, Obersteiner M, Evangelopoulou S, De Vita A, Winiwarter W, viron 2018;64:5e14. https://doi.org/10.1016/j.trd.2017.07.012.
Witzke H-P, et al. EU Reference Scenario 2016 Energy, transport and GHG [27] Moore MA, Boardman AE, Vining AR, Weimer DL, Greenberg DH. “Just give me
emissions Trends to 2050. n.d. https://doi.org/10.2833/9127. a number!” Practical values for the social discount rate. J Pol Anal Manag
[12] Capros P, Kannavou M, Evangelopoulou S, Petropoulos A, Siskos P, Tasios N, 2004;23:789e812. https://doi.org/10.1002/pam.20047.
et al. Outlook of the EU energy system up to 2050: the case of scenarios [28] Rambaud SC, Torrecillas MJM. Some considerations on the social discount
prepared for European Commission's “clean energy for all Europeans” pack- rate. Environ Sci Pol 2005;8:343e55. https://doi.org/10.1016/
age using the PRIMES model. Energy Strateg Rev 2018;22:255e63. https:// j.envsci.2005.04.003.
doi.org/10.1016/j.esr.2018.06.009. [29] Zhuang J, Liang Z, Lin T, De Guzman F. ERD working paper series no. 94.
[13] Osborne MJ, others. An introduction to game theory, vol. 3. New York: Oxford Theory and practice in the choice of social discount rate for cost-benefit
university press; 2004. analysis: a survey. 2007.
[14] Nash J. Non-cooperative games. Ann Math 1951;54:286e95. [30] Harrison M. Valuing the future: the social discount rate in cost-benefit anal-
[15] Nash J. The bargaining problem. Econometrica 1950;18:155e62. ysis. Melbourne: Productivity Commission; 2010.
[16] Aalborg University. Department of Development and Planning. EnergyPLAN | [31] Steinbach J, Isi F, Staniaszek D. Discount rates in energy system analysis
Advanced energy systems analysis computer model n.d.. https://www. Discussion Paper. 2015.
energyplan.eu/. [Accessed 12 February 2021]. [32] García-Gusano D, Espegren K, Lind A, Kirkengen M. The role of the discount
[17] Fischer R, Elfgren E, Toffolo A. Towards optimal sustainable energy systems in rates in energy systems optimisation models. Renew Sustain Energy Rev
Nordic municipalities. Energies 2020;13. https://doi.org/10.3390/ 2016;59:56e72. https://doi.org/10.1016/j.rser.2015.12.359.
en13020290. [33] Lazard’s Lazard. Levelized cost of energy analysisdversion 14.0. 2020. New
[18] Energimyndigheten. Antal verk och installed effekt per kommun. 2003. n.d, York, US.
https://pxexternal.energimyndigheten.se/pxweb/sv/Vindkraftsstatistik/ [34] Hartman RS, Doane MJ. Household discount rates revisited. Energy J 1986;7.
Vindkraftsstatistik/EN0105_4.px/. February 12, 2021. [35] Schleich J, Gassmann X, Faure C, Meissner T. Making the implicit explicit: a
[19] Radio Sveriges, Norrbotten P4. Storsatsning på vindkraft i Piteå sk€ argård. look inside the implicit discount rate. Energy Pol 2016;97:321e31. https://
Sverige Radio 2006. https://sverigesradio.se/sida/artikel.aspx? doi.org/10.1016/j.enpol.2016.07.044.
programid¼98&artikel¼1022396. May 15, 2019. [36] Copiello S, Gabrielli L, Bonifaci P. Evaluation of energy retrofit in buildings
[20] Swedish government. Sweden's fourth national energy efficiency action plan. under conditions of uncertainty: the prominence of the discount rate. Energy
Stockholm; 2014. 2017;137:104e17. https://doi.org/10.1016/j.energy.2017.06.159.