Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

Impact of operation strategies of large scale battery systems on distribution MARK

grid planning in Germany
⁎
Matthias Rescha,b, , Jochen Bühlera, Mira Klausena, Andreas Sumperb
a
Reiner Lemoine Institut gGmbH, Ostendstraße 25, 12459 Berlin, Germany
b
Centre d'Innovació Tecnològica en Convertidors Estàtics i Accionaments (CITCEA-UPC), Departament d'Enginyeria Elèctrica, Universitat Politècnica de
Catalunya, ETS d'Enginyeria Industrial de Barcelona, Av. Diagonal, 647, Pl. 2. 08028 Barcelona, Spain

A R T I C L E I N F O A BS T RAC T

Keywords: Due to the increasing penetration of fluctuating distributed generation electrical grids require reinforcement, in
Grid planning order to secure a grid operation in accordance with given technical specifications. This grid reinforcement often
Distribution grid leads to over-dimensioning of the distribution grids. Therefore, traditional and recent advances in distribution
Large scale batteries grid planning are analysed and possible alternative applications with large scale battery storage systems are
Community storage
reviewed. The review starts with an examination of possible revenue streams along the value chain of the
Primary frequency control
German electricity market. The resulting operation strategies of the two most promising business cases are
discussed in detail, and a project overview in which these strategies are applied is presented. Finally, the impact
of the operation strategies are assessed with regard to distribution grid planning.

1. Introduction with wind power plants in the north and photovoltaic systems in the
south [10]. This, and the fact that the power feed-in of DG is not
The energy system in Germany is currently changing. In the past, necessarily simultaneous to the local load demand, results in a
electrical energy was injected by large power plants into the transmis- transformation process of the distribution grids. Formerly they were
sion system (220 kV and 380 kV) to cover long distances. It was then characterised by the consumption whereas now the reverse power flow
delivered to costumers via distribution (smaller) grids (1–110 kV). becomes increasingly common. This means that in some moments of
Since the German Federal Government decided to withdraw from the the year there is a power flow from the distribution grid to the
nuclear energy programme and to reduce the greenhouse gas emissions transportation grid [11]. As German electricity grids are planned to
in order to mitigate climate change, the expansion of renewable energy work uni-directional with a power flow from high to low voltage levels
sources was subsidised by introducing the German Renewable Energy this could lead to several problems. For example the protection concept
Act (EEG) in 2000. This led to a tripling of the share of renewable is designed such as to work for an uni-directional power flow and may
energy in the German electricity mix from 7% in the year 2000 to 25 % not work in a bi-directional way [12]. Furthermore, power quality
in the year 2013 [1]. As a consequence, the sinking levelised cost of issues can arise. In some grids the maximum possible PV penetration
electricity (LCOE) of renewable energy sources (RES) led to grid parity. rate is reached as DG are often installed in rural grids [13]. Therefore,
This trend will probably continue as the German Federal Government an additional installation of DG is often followed by grid reinforcement
committed itself to a RES ratio of 80% of the gross electricity in order to solve over-voltage and equipment over-loading issues. The
production in the year 2050 [3]. In contrast to conventional power drawback of this traditional grid planning procedure is large invest-
plants, RES are mainly realised as distributed generators (DG), as ment in infrastructure with a low utilisation rate. Historically, network
defined by [4,5]. Due to their relatively small installed nominal power extension planning has been based on maximum load scenarios, but in
they are mainly connected to the distribution grid at medium voltage the case of a high penetration with DG the grid is dimensioned to deal
(MV) and low voltage (LV) levels [6,7]. For example, 80% of photo- with maximum generation [14]. In Germany, the number of hours in
voltaic (PV) power plants in Germany are connected to the LV grid [8]. which PV-systems feed more than 90% of their nominal power into the
Due to this, the nominal DG power installed in the distribution grid grid is below 100 h a year [15]. Due to this, traditional grid planning
surpassed the power installed in the transportation grid in 2010 [9]. may cause inefficient grid operation and higher grid utilisation fees that
Furthermore, the DG are distributed very inhomogeneous in Germany have to be borne by the general public (cost increase of 9,2% from 2008

⁎
Corresponding author at: Reiner Lemoine Institut gGmbH, Ostendstraße 25, 12459 Berlin, Germany.
E-mail address: [email protected] (M. Resch).

http://dx.doi.org/10.1016/j.rser.2017.02.075
Received 29 April 2016; Received in revised form 1 December 2016; Accepted 21 February 2017
Available online 17 March 2017
1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

to 2014) [16]. As in [17] predicted, this will lead to a linear cost transformers [17] sets the maximum loading to 120%. For all other
increase for DG induced grid reinforcement due to over-voltage and components and scenarios it is set to 100%. Nevertheless, the
over-loading issues of 331 EUR/ kW until 2030. The cost can be maximum loading of MV/ LV transformers depends not only on the
designated to different voltage levels (400 V:13%/1–36 kV:29%/60– profile but is also not consistent in the literature: it ranges from 150%
380 kV: 58%). Therefore, the impact of different operation strategies of for oil immersed transformers only [28,29] to 120% [30,31] and 100%
microgrids [18], electrical vehicles [19] and residential storage systems [17] for all kind of transformers in the case of a RPF caused by PV
[20,21] to increase the hosting capacity of DG in distribution grids have systems.
been analysed by the authors. Although [22] provides an overview of Voltage characteristics of electricity in distribution grids are defined
(large scale) energy storage technologies suitable for wind power in [26]. The most important restrictions are that the frequency has to
application, the implications of the operating strategies as for example be kept at 50 Hz ± 1 Hz and the 10-min RMS average of the voltage at
voltage control for distribution grid planning have not been analysed in the point of common coupling (PCC) has to be kept with in an interval
detail. Extending the previous work of the authors, this paper gives an ± 10% of the nominal voltage. To ensure this two technical specifica-
overview and evaluates alternative possibilities to traditional DG tions for DG quantify the permitted voltage rise of 2% in the MV [32]
induced grid extension with large scale battery storage systems and of 3% in the LV, respectively [33]. These technical specifications
(BSS). As in most cases, this alternative turns out not to be profitable, apply if the MV or the LV are calculated separately, otherwise these
if the BSS's only purpose is to mitigate traditional grid extension [14] thresholds don‘t have to be considered. Furthermore, all generators
additional revenue streams have to be taken into account. Therefore, connected to the electrical grid have to comply with the specifications
the objectives of this paper are to review additional applications for of [34,35] and [36], respectively. Furthermore, the technical note [37]
BSS in the German electricity market in order to combine them with has to be considered for BSS connected to the LV.
the task of mitigating grid extension caused by DG and evaluate the The technical restrictions for over-voltage and over-loading are
impact of the resulting operation strategies on traditional and new commonly used to determine the hosting capacity, as defined in [38], to
approaches of distribution grid planning. integrate DG into existing grids. An exhaustive international overview
The paper is structured as follows: In Section 2 the legal framework of the main technical issues limiting the hosting capacity for DG of
for the operation of distribution grids in Germany and the challenges distribution feeders is given in [39].
that arise with integration of high shares of DG are described briefly.
Section 3 covers traditional distribution grid planning and in Section 4 2.2. Challenges and solutions for electrical grids with fluctuating
new grid reinforcement planning methods are presented. A brief feed-in of renewable energies
overview of different BSS applications and their possible profit margins
for the German energy market is presented in Section 5. In the same In this subsection the challenges that arise from the integration of
section, the implementation of large scale battery systems in distribu- high shares RES into the electrical grid are discussed. The increasing
tions grids is discussed. The focus lies on BSS that apply self- penetration of DG has, among other issues, led to the following [39–
consumption maximisation and primary control reserve, due to their 41]:
economical relevance, as well as the possible impact on the grid For distribution grids in particular:
planning. Finally, the conclusions are summarised in Section 6.
• Thermal over-loading of network equipment
• Voltage rise
2. Legal framework, arising challenges and possible • Increased fault levels, especially for MV grids
solutions for DG and BSS connected to distribution grids in • Power quality issues
Germany • Impact on grid protection due to RPF
• Effect on the operation of voltage regulators and tap changers
2.1. Legal framework for the operation of distribution grids because of RPF
• Impact on grid losses
According to the German Energy Act (EnWG) Section 14(1) [23]
the grid operators are legally bound to ensure a safe and stable energy For the whole electrical system:
supply. Especially the power quality issues of over-loading of cables
and transformers as well as over-voltage are of major interest. The • Increased demand of control power
parameters that should fulfilled regarding over-loading of transformers • Increase of transmission line bottlenecks
and LV-cables are defined in DIN EN 60076-2:2011 [24] and DIN VDE • Decreasing spinning reserve
0276–603 [25], respectively. Table 1 shows the load factors of the rated
apparent power Sr for different components according to [17] under The most important challenge in distributions grids on an interna-
normal operation conditions that are defined in [26]. For the heavy tional level is due to over-voltage issues [39]. In Germany for example,
load flow (HLF) and reverse power flow (RPF) different maximum load 80% of the grid reinforcement is due to over-voltage issues in
factors apply. This is due to the different shape of the profiles in both distribution grids [42]. Besides grid reinforcement, ancillary services
cases. Furthermore, the (n-1)-criterion as defined in [27] and further have to be provided by generators and loads to cope with these issues.
specified in [17] applies for MV-cables and HV/ MV transformers for These services are defined in [43] and classified as follows for normal
the load case. In the case of a HLF for MV-cables and HV/ MV operation conditions:

Table 1 • Frequency control

Equipment load factors [17].
• Voltage control
• Remote automatic generation control
•
Equipment Load factor of Sr Load factor of Sr
Heavy load flow Reverse power flow Grid loss compensation

LV-cable max. 100% max. 100% All these ancillary services can be provided by DG and in particular
MV/LV tran. max. 100% max. 100%
by BSS [44]. Therefore, the technical and economic applications of BSS
MV-cable max. 60% max. 100%
HV/MV tran. max. 60% max. 100% are analysed hereafter in order to supply ancillary services and as an
alternative to traditional grid reinforcement.

1043
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Table 2
Diversity factors for generators connected in MV or LV [7,17,53,54].

Wind PV BM Water

HLF 0 [17] 0 [7,17,53,54] 0 [17] 1 [17]

0.6 [7]
RLF 0.95 [7] 0.85 [17,53,54] 0.98 [7] 1 [17]
1 [17] 0.89 [7] 1 [17]

for Germany sets the scenarios closer to the reality. For loads, this
scenario parameter is called coincidence factor and is defined in [46] as
the average power absorbed related to the installed power. For
generators, this factor is referred to as diversity factor by [52], and is
defined as the quotient of the actual and the installed capacity. To
quantify the coincidence and the diversity factor taking into account
the simultaneity of generation and consumption, several studies have
been conducted [7,17,53,54]. The diversity factors of [7] apply for ten
generators of the same type. The same study presents that diversity
factor differs if the correlation between the generators is taken into
Fig. 1. Distribution grid schematic [202,17] (adapted).
account. The results are listed in Tables 2 and 3.
The coincidence and diversity factors all apply to the maximum/rated
power of the generators and loads. In case of PV this factor refers to the
installed module power PSTC [53,54]. In the reverse power flow case the
3. Traditional distribution grid planning factor for the load of the MV is higher, as higher blending of the stochastic
behaviour of the loads is taken into account. Some bigger costumers/loads
Although, there are different guidelines for distribution grid have their own secondary transformer and are connected directly to MV (C.
planning on a national [45] and an international level [46], as well load). The maximum power of these loads can be assumed as 40% of the
as recommendations like [47], every DSO has a different planning rated apparent power Sr , t of the secondary transformer [55]. Based on
process because of the different characteristics of each distribution experience, these simple worst case parameters cover all possible grid
grid and DSO [48]. To standardise the different planning ap- states. These worst case scenarios are therefore commonly used, e.g. in
proaches a study was conducted that summarises the methodology [56,57,17], as this method provides a high level of reliability without
of 17 DSO covering more than 50% of all distribution grids in measurements in the LV [45]. The likelihood of these extreme grid states
Germany [17] and which can be regarded as the state-of-the-art however, is not considered with this practice, and may never occur in reality
approach. Fig. 1 describes the conventional distribution grid plan- [58]. Furthermore, no time interdependencies of the assets are considered.
ning schematically. As a consequence, the distribution grids tend to be over-dimensioned. Thus
One problem of this approach lies in the input data, since the LV [59] and [46] claim that new planning approaches should be taken into
load is usually not measured and has to be estimated. The estimated LV account as they may use infrastructure more efficiently [47], as well as
load may be gained from the (measured) annual maximum load of the avoid redundant investments and minimise O & M costs [60]. There are
secondary transformers [49], the rated power of these transformers plenty of different approaches to come to a more realistic assessment of the
[50] or structural data as the degree of electrification or population scenario parameters, as for example [55,52,61].
density [49,45]. Also, approaches employing combinations of these In general, there is a wide field of different new planning
datasets are possible and described in [17]. On the generation side the approaches for different applications which are analysed in Section 4.1.
rated power of the generators are usually well known and published
[51].
To evaluate whether a certain threshold is reached (as described in 3.2. Grid reinforcement methodology
Section 2) a power flow calculation is conducted in which the power of
the load and the generator are adjusted to certain worst case scenarios, Hereinafter the methodology of grid reinforcement for distribution
specified in Section 3.1. If a threshold is passed, the grid will be grids, especially for low and medium voltage grids, is described. The
reinforced according to the methodology described in Section 3.2. methodology is depicted in the figures for radial grid structures in the
LV and for open ring structures in the MV. Nevertheless, these
3.1. Assumed scenarios - worst case parameters methodologies are transferable to other grid topologies and can be
considered as state-of-the-art in Germany [17]. As described before,
Distribution grids are traditionally planned in a deterministic triggers for grid reinforcement are either local over-voltages or over-
manner [46]. The traditional scenario to conduct a power flow only loadings of a cable or a transformer. First, the over-loading measures
considers maximum demand, whereas the generation is assumed to be are implemented, then another load-flow is conducted. If there are still
constant. As aforementioned, the higher penetration rate of DG leads to over-voltage problems in the grid, the measures to solve these apply.
two worst case considerations: the heavy load flow (HLF) and the
reverse power flow (RPF) scenario. On an international level they are Table 3
parametrised according to [46]: Coincidence factors for loads connected in LV and MV [17,55].
1) Heavy load flow: Max load; no generation.
Load (LV) Load (MV) C. load (MV)
2) Reverse power flow: Min load; max generation.
These extreme parameters do not consider the time variability of HLF 1 [17] 1 [17] 1 [55]
demand and generation. Thus, a simple probabilistic determination of RPF 0.1 [17] 0.15 [17] 0.5 [55]
the worst case scenario parameters that covers all possible grid states

1044
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

For example [62,63,31] regard low voltage exclusively, whereas [64]

focuses only on the medium voltage and [30] considers both voltage
levels. Another difference in [62,31] is that the new parallel line is
installed from the secondary transformer to the distribution cabinet
closest to the critical node within the feeder.
According to [17,62,31] all new lines are supposed to be under-
ground cables, instead of overhead lines due to the higher acceptance of
the general public. For an easier automation the reinforcement
equipment is standardised but differs from case to case as shown in
Table 4.
According to [13] who conducted a statistical analysis of distribu-
tions grids in southern Germany, the NAYY 4x 240 mm2 is the most
Fig. 2. LV grid reinforcement via a parallel line. commonly used cable type in LV (36% in rural grids, 84% in villages
and 38% suburban grids) and is used twice as often as any other cable
type.

Methodology for low voltage grids:

4. New planning methods and definitions for BSS
As depicted in Fig. 2, an over-voltage is solved by installing a
parallel cable (type see Table 4) from the distribution substation to the
4.1. New planning methods for integrating DG and BSS in
next distribution cabinet over 2/3 of the line length. A critical over-
distribution grids
loading of a line is solved by installing a parallel line till the next
distribution cabinet, starting to search from half of the line on.
The aim of the reviewed studies in this section is to determine,
If more than one line is affected, as shown in Fig. 3, all affected lines
besides other network parameters, the optimal number, location and
are divided at the distribution cabinet that lies closest behind one half
size of DG and BSS units. This is achieved by optimising the total
of the line. The lines of the second half are connected to a new
capital expenditures (CAPEX) and operational expenditures (OPEX)
secondary substation. The rated apparent power Sr , t of the additional
MV/LV transformer is the same as the one that was formerly feeding including DG and BSS. Several objectives have been pursued via this
the entire LV-grid. If there is an over-loading in a transformer and its optimisation of DG integration in distribution grids. Some of the most
apparent power Sr , t ≤ 400 kVA , it is replaced by the next bigger common objectives are: minimisation of energy losses, maximisation of
standard transformer (630 kVA). If the over-loading is not solved, a DG capacity or energy via sizing and allocation of DG, minimising
parallel 630 kVA transformer is installed. curtailment losses, minimising costs, as well as the minimisation of the
grid reinforcement cost associated with DG [65]. The planning process
Methodology for medium voltage grids: can be described as a non-linear mixed integer optimisation problem.
Similar to the LV a parallel line is installed in the case of over- There are several comprehensive reviews for new distribution grid
loading or over-voltage. In case of over-voltage the length of the new planning approaches. While [66–68] describe and classify the planning
line is 2/3rd of the length of the affected feeder, whereas for over- approaches generally, [65,69,70] concentrate on DG integration.
loading the parallel line is installed between the primary substation and Hereafter, the criteria and definitions as well as the three-level tree-
the DG that causes the trouble (see Fig. 4). It applies for both measures structure according to [66] are used to classify the planning methods
that no secondary substations are installed on the parallel MV line used in a selection of reviewed studies listed in Table 5.
which is connected to the bus bar of the primary substation. At the According to the first level of the tree-structure, all methods can be
connection points an additional breaker is installed in the affected divided into models with or without reliability considerations. In
feeder. planning models without reliability features the grid is operated under
If the parallel cable does not solve the issue, a new MV ring is operational constraints. The aims of planning optimisations are mini-
installed according to Fig. 5. By this measure the critical part of the mising CAPEX of substations, feeders or feeder branches (assets),
affected open MV ring is transferred to two uncritical open MV rings by minimising the costs of capacity upgrades of the existing facilities, as
separating the DG that causes the problems with a parallel MV line. well as minimising the OPEX and the energy losses.
The costs for the earthworks apply only once, as it is assumed that both In the second level, the models may or may not include uncertainty
lines share the same trench. considerations. In contrast to deterministic planning, uncertainty
If the HV/MV transformer is over-loaded it is replaced with a 40 models consider the unpredictability of future load demand and
MVA transformer. If the over-loading still remains a parallel 40 MVA generation at the design stage. The reliability considerations may be
transformer for the same feeder is installed. In case all the aforemen- considered in the planning and can be incorporated either under
tioned measures do not solve the problems, a new primary substation is normal conditions or under contingency conditions. To include the
installed as depicted in Fig. 6. In this case, the placement of the new maximum network reliability under normal conditions, a reliability
substation and new breakers is done manually in order to solve all objective function minimising the expected outage cost or expected
occurring issues in the MV-grid manually. annual non-delivered energy is added to the other objective functions.
Other studies [62–64,30,31] suggest slightly different approaches. In order to include predefined fault/ contingency conditions an
objective function similar to the aforementioned reliability objective

Table 4
Standard equipment for grid extension.

Equipment dena [17] Stetz et al. [62] Idlbi et al. [64] Ackermann et al. [30]

2 2 2
LV-cable (NAYY) 4×150 mm (3×150; 3×240) mm 4×150 mm 4×150 mm2
MV/LV tran. (Sr , t ) 630 kVA (400; 600; 800) kVA (400; 600; 800; 1000) kVA 630 kVA
MV-cable (NA2XS2Y) 3×1×185 mm2 – – 3×1×240 mm2
HV/MV tran. (Sr , t ) 40 MVA – – –

1045
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Fig. 5. MV grid reinforcement via an additional MV ring.

Fig. 3. LV grid reinforcement via an additional secondary substation.

respective power flows. Whereas, [72] uses MILP to calculate

function is employed. the optimal size and location of feeders and substations over the
The third level categorises all types of optimisation models depend- planning horizon of 10 years.
ing on the type of (decision) variables and objectives. There are (a) 2. Mixed-integer non-linear programming (MINLP)b
mixed-integer (b) discrete and (c) continuous models. Commonly, MINLP refers to optimisation problems with continuous and
integer variables in distribution system planning problems are used discrete variables and a non-linear objective function and/or
for decisions on whether or not new assets are installed or existing non-linear constraints.
equipment is replaced or extended. Discrete variables are usually used In [73] a MINLP is used to decide whether to invest in DG
for the dimensioning of the equipment, whereas continuous variables and/or purchase power from the main grid and invest in feeders
are generally used for voltages and power flows. In mixed-integer and substations in case of future load growth. Another approach
models, all three types of variables can be optimised. Discrete and is used by [74], who formulated the MINLP as a TRIBE particle
continuous models on the other hand are restricted to discrete and swarm optimisation and ordinal optimisation with the aim of
continuous decision variables, respectively. All reviewed studies are minimising total costs by optimal allocation of DG. Reactive
categorised within these three models and listed in Table 5 including capabilities of different DG and uncertainty in load demand and
their type of solution strategy. The different solution strategies are generation has been analysed. However, BSS have not been
discussed in [66] with more detail. For reasons of conciseness the considered. A multi-objective optimisation using MINLP in
various methods have been denoted with indices, which are used in order to find a trade-off between minimising the investments
Table 5. and the emission of pollutants, taking into account uncertain
market prices, has been presented by [75].
(a) Mixed-integer models 3. Bender's decomposition (BD)c
The mixed-integer models are the most common ones. They In this algorithm the mixed-integer model is separated into
combine binary decision variables (1(Yes), 0(No)) with a set of two discrete models: the discrete ’relaxed master problem’ and a
continuous and discrete variables. quadratic ‘sub-problem’. First, the master problem is solved to
1. Mixed-integer linear programming (MILP)a decide on investments in new equipment. Secondly, the quad-
MILP is a two-step approach. In the first step, an initial ratic sub-problem is solved to optimise the power flow in order
solution is determined by solving a linear problem, where all to minimise the operational costs.
variables are treated as continuous variables, usually using the A long-term multi-stage model has been presented by [76]
simplex algorithm. In the second step, successive searches are and [77]. This model uses new-path and fencing constraints to
performed to obtain better solutions for the integer variables. reduce the complexity of the solution space. This grid expansion
For example in [71] MILP is used to determine the achiev- planning method minimises investment costs for growing load
able gross margin in the different electricity markets for BSS and demand by including DG, similar to [73].
its resulting operation, as well as for the determination of the
storage redispatch and DG curtailment measures and their

Fig. 4. MV grid reinforcement via a parallel line. Fig. 6. MV grid reinforcement via an additional primary substation.

1046
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Table 5
New distribution grid planning approaches with DG integration.

without reliability with reliability under with reliability under

normal conditions contingency conditions

deterministic uncertain deterministic uncertain deterministic uncertain

b c c a,b b,z y e e d,z e,z

mixed integer [73] , [77] , [76] , [72] [75] , [91] [85] , [83] [82] [74]
[78]d, [86]f, [63]e, [71]a,d [92]y [84]e [79]d,z – [80]d,z
continuous [72]h, [71]h [89]g,z – – [81]d –

a b c d e
MILP; MINLP; BD; GA; PSO; f ES; g
DP; h
NLP; y
possibilistic; z probabilistic.

4. Genetic algorithm (GA)d (c) Continuous models

Inspired by natural evolution processes in genetic algorithms In continuous models the considered variables have to be
generations of individuals exist. Simulating the evolutions of continuous and thus the need of discrete decision variables is
individuals by emulating the process of selection, mutation and eliminated.
recombination of genes, the reproduction is based on fitness 1. Dynamic programming (DP)g
functions preferring the best individuals. GA can be used for Dynamic programming allows to represent the ever-chan-
different purposes in distribution grid planning: In [78] it is ging nature of the planning process. This is realised by model-
used to find the optimal grid topology. In [71] the GA is used for ling the states of the network in nodes with certain states. These
BSS allocation and calculation of grid reinforcement measures. states can change in time with every investment in grid
The optimal trade-off between traditional grid expansion and reinforcement and are based on the former state. In [89] this
implementation and/or the energy purchase of DG is considered method is used to realise a long-term planning (10 years) for the
in [79–82]. optimal sizing, allocation and most important the timing of
5. Particle swarm optimisation (PSO)e investment in DG based on measured values (current and
PSO is another evolutionary algorithm that simulates in- voltage).
dividuals (particles) in a swarm and their social behaviour. A 2. Non-linear programming (NLP)h
vector is used to locate every particle and its velocity in the NLP is a numerical method, which only accepts continuous
swarm. The population of particles searches for the optimal variables. The most common application for NLP in the context
solutions using the individual experience of the particles and of distribution grid planning is AC optimal power flow (AC-
sharing it with the others. The swarm can also return to OPF), as used in [71] to minimise active power redispatch for all
promising regions found before. Generally it is used to allocate DG and BSS. NLP is applied by [72] to determine the optimal
DG [83,84,74] and/or BSS [85], or on-load tap-changer [63]. capacities and production of the DG.
PSO might also be used to calculate the minimal reactive power
output of DGs to solve over-voltage problems [63]. Qualitative evaluation
6. Expert system (ES)f The biggest advantage of these models are, that no linearisation is
Expert systems are knowledge-based systems, that try to required making it a good choice for extension planning purposes of
emulate the decisions a human would make. Besides heuristic large scale distribution grids. The drawbacks are, besides the large
rules a broad data basis like GIS-Data, economic data from computational effort [60], that these models are badly suited for
asset-management databases as well as the grid topology and greenfield considerations [90].
measured data are combined for this purpose to create a semi- Furthermore, all the methods might be either deterministic or
automatic grid planning process [86]. consider uncertainty in the model. The uncertainty can be considered
by using a possibilisticy approach as used in models that apply a fuzzy
Qualitative evaluation total installation and operational cost or a fuzzy non-delivered energy
Mixed-integer linear models allow a high degree of generalisation. as objective function. A multi-objective optimisation based on fuzzy
Nevertheless, in order to optimise real grids, non-linear characteristics logic has been presented by [91], who uses a Bellman-Zadeh algorithm
like cost functions and grid characteristics have to be linearised. to analyse a wide range of technical, economic and environmental
Consequently, the optimal solution is not necessarily the best for the criteria to find optimal allocation of DG in distribution grids.
real system, due to the simplifications [87,88]. Another approach to handle uncertainty is called probabilisticz
approach. In this model the uncertainty is calculated by applying a
(b) Discrete models probability distribution function. The power generation or the size of
In these models, discrete and binary variables (yes/ no) are the DG is a common example for a probabilistic application.
used in the objective function formulation to deal with the decision As presented in Table 5, deterministic approaches without relia-
on location and size of the network facilities. bility considerations show the highest variety of numerical and
evolutionary methods, and are the most commonly used. In studies
Qualitative evaluation that implement reliability considerations evolutionary algorithms seem
Discrete models allow the determination of the timing of reinforce- to be the predominant method, because of their advantage to optimise
ment measures for long term planning, but only discrete variables are several criteria at the same time.
allowed. Generally, the same restrictions for large scale systems apply In most of the cited studies the DSO is at the same time the owner
as for mixed integer models due to high number of possibilities [66]. To of the DG, BSS or OLTC-transformer and can decide on the allocation
the authors knowledge discrete models are not applied for DG and/or the operating strategy of the equipment
integration in distribution grids, as no work has been published on [73,74,77,76,78,75,91,89,85,83,84,82,79–81]. Only few works consid-
this topic in the public domain. er that the equipment might be privately owned and operated, as is the

1047
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

case in Germany [72,86,63,71,92]. In Germany, due to unbundling the (b) Grid supportive
DSO is normally not the owner of the DG or BSS and therefore has only This characteristics describes the behaviour of the BSS of
very little or no influence on the location. Furthermore, the volatile actively stabilising the grid that goes beyond the minimal pre-
character of the DG, as well as the stochastic behaviour of loads and the requisites described before. It has a local component as some
possible participation of DG, BSS and loads at the energy market, lead issues like over-voltage and over-loading have to be solved locally.
to extreme scenario parameters. Consequently, the grid is over- Over-voltage may be solved with active or/and reactive power
dimensioned, if the conventional planning based on worst case control as addressed for instance by [96]. The market incentive
scenarios is applied. The over-sizing problem remains with the programme from the German Federal Government and the state-
presented new grid distribution planning methods as long as extreme owned KfW banking group is coupled to several technical require-
scenarios are used, even if the systematic approach of the new methods ments. The most important measure in this context is the limita-
eliminate the uncertainty of manual planning. The problem can be tion of maximum feed-in power of the PV storage system to 60% of
solved by applying possibilistic or probabilistic methods. Probabilistic its nominal power at the point of common coupling [98].
algorithms use probability density functions for loads and generation to (c) System compatible
quantify the likelihood of grid states as for example very rare loading Analogue to a grid compatible behaviour a system compatibility
situations and can derive the reliability of the electrical power supply. is given with the fulfilment of the minimal requirements of the BSS
The main drawback is that high quality time-series of the grid to ensure a safe operation of the whole electrical system. In this
participants are needed to generate the probability density functions, case the contribution to the spinning reserve, as well as the
which are often not available in LV grids. This applies especially for the provision of ancillary services as for instance black start capability
active power flow of BSS as their operation strategy depends on the and frequency control play an important role. Some of these
business case which might depend on the energy market for instance. services, like the provision of primary frequency control, are
Furthermore, the reactive power flow of the BSS, depends on other remunerated whereas some, such as the provision of spinning
network participants and on the current grid state. Consequently, to reserve or active power reduction in case of over-frequency, are not
generate realistic time-series existing interdependencies in the dis- [96].
tribution grid as well as business case related issues have to be (d) System supportive
considered. These time series can be used as an input for any planning A BSS can be considered system supportive, if it leads to
optimisation method mentioned above and should be an improvement greater flexibility of the electrical system. The operation of the BSS
to traditional worst case considerations. is then optimised to minimise local issues as described for the grid
Several studies exist combine grid planning with DG and take the supportive behaviour and at the same time to provide services for
active power control of large scale batteries for peak shaving into the whole electrical system. An example may be the provision of
account [85,83,14,93]. Nevertheless, from the studies mentioned above reactive power to reduce local over-voltage issues and the provi-
only [85] and [93] consider reactive power control, even though [94] sion of active power to provide frequency control and/or spinning
highly recommends further studies on this issue. This is due to the fact reserve.
that reactive power control from BSS is as a very easy and cost-effective
way of voltage control which is independent from the state of charge of 5. Overview of large scale battery systems in distribution
the battery. grids

4.2. Definitions of behaviours of BSS Large scale battery systems are not clearly defined. They may be
defined by their type of operation, as in [99,100]. In [99] large scale
As stated before, BSS may provide active and reactive power. The BSS are delimited from small scale BSS, if they supply peak levelling
application dependent power flow may either lead to less or to services and are grid connected or if power-quality control applications
additional grid reinforcement cost [17]. In this section different system are applied. A more specific definition of the application of large scale
behaviours and the criteria of the BSS in order to quantify their impact BSS is given by [100], who distinguishes between energy related or
on the distribution grid planning are defined. In this study the term power related applications. In energy related applications, the storage
system refers to electrical systems, whereas the heat and transport is charged and discharged during several hours, reaching one cycle a
sector are excluded. Every BSS may be categorised in one or several of day. In contrast to this, for power applications the BSS is cycled several
the four categories mentioned hereafter [95,96]: times a day and discharged and charged in shorter periods (typically
seconds and minutes). The type of application directly affects the range
(a) Grid compatible in which the rated power range of the BSS tends to be and might be
If the minimal technical requirements in regard to quality, used as an indication, as listed in Table 6 according to [101].
reliability and safety imposed by the DSO are fulfilled by the BSS, it In Section 5.1 the market potential of large scale BSS in German
can be considered as grid compatible. In the near future operators distribution grids according to the definition mentioned above is
of DG will need to prove this behaviour via certificates to the DSO. estimated. The operation strategies of the two most promising business
Based on the criteria for to PV systems, possible future criteria cases are analysed in Sections 5.2 and 5.3 and the impact of the
which have to be proven by the BSS, are [97]: operation strategies is concluded in Section 5.4.
(i) Short-circuit current capability, (continuous) current carry-
ing capacity ampacity and switching capacity of the main 5.1. BSS applications and German energy market
components
(ii) Active power feed-in In broad terms, there are two ways to gain monetary benefits along
(iii) Active power concept the electricity value chain with existing BSS applications in the German
(iv) Network disturbances like rapid voltage drops, long-term electricity market: first, revenues received by the storage owner or
flicker, harmonics and interharmonics operator and second, cost reduction or avoidance by the storage owner
(v) Fault ride through or operator [102]. Generally, revenues can be achieved through existing
(vi) Contribution to the short circuit current markets and bilateral contracts. Cost reduction or avoidance on the
(vii) Static provision of reactive power other hand is highly based on individual use cases. Some important
(viii) Conditions for connecting and protection concept for dis- application analyses have been summarised for the German electricity
connecting the system market in [103–106] and are shortly presented in the next sections

1048
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Table 6 of black start units plus a concept for the restoration of supply in
Energy and power related applications for BSS [101]. their control area. The black start capability is not explicitly
defined in the Transmission Code. The requirements for the type,
Application Nominal power P
scope and remuneration are negotiated bilaterally.
Energy related: (iii) Redispatch:
Peak shaving 0.1–10 MW In many areas in Germany, transmission capacities are not
Load levelling 1–100 MW
keeping pace with the changing feed-in and off-take infrastruc-
Energy arbitrage 50–500 MW
Power related: ture. In order to ensure security of supply, TSOs with the help of
Frequency control 1–30 MW DSOs take redispatch measures, adjusting feed-in from parti-
Voltage regulation 1–30 MW cular generating and storage facilities [113]. A transparent
Power quality regulation 1–30 MW market for redispatch does not exist. The selection of generators
Bridging power 1–30 MW
for redispatching is based on their location in the network, their
generation form and their size, which determines either the cost-
based (where the adequacy of costs is regulated) or market-
together with their potential benefit estimations: based (based on individual bids submitted by the generators)
redispatch [114].
(a) Market revenues (c) Cost reduction or avoidance
(i) Power exchange markets: (i) Uninterrupted power supply (UPS):
As electricity is a homogeneous commodity and the majority of Large and long power cuts (>3 min ) arise relatively arbitrarily
the power supply must be consumed at time of production, in Germany. However, voltage dips (<1 min ) as well as short
electricity prices show a high volatility. In addition, the short-term interruptions (<3 min ) occur 10–100 times per year [114].
demand is not very price elastic [107]. These circumstances allow Therefore, depending on the specific outage times and individual
inter- temporal arbitrage transactions at the EPEX-Spot (day- power quality needs (e.g. voltage, frequency, harmonics), a UPS
ahead and intraday market). Arbitrage contains purchases of system can consist of a BSS in combination with a generation unit
electricity in times of low energy prices (off-peak prices) and sales like a diesel or gas generator or of a battery only [106].
of electricity when prices are comparatively high (peak prices) (ii) Balancing group management (BGM):
[108]. The attractiveness of the application depends on price with the liberalisation of electricity markets in Europe and
spreads and the frequency of price spreads in these markets. On Germany, the balancing group system was established.
the day-ahead market, 24 h single contracts and diverse block Accordingly, each producer or consumer must belong to a balance
contracts are traded for the next day via a daily static auction. The group and all balance groups must be levelled at a quarter-hourly
intraday market starts shortly after the day-ahead market (trades basis. The German TSOs are liable for determining and settling the
for the following day start at 3 pm and end 30 min before the amounts of balancing energy in their control area, using a common
actual physical delivery of the respective contract) and is organised symmetric imbalance price for each 15-min time period (German:
by continuous trading. regelzonenübergreifender einheitlicher Bilanzausgleichsenergiep
(ii) Control reserve markets: reis, reBAP) [115]. Consequently, a BSS can optimise the indivi-
A stable operation of the power supply system at a system dual energy balancing costs.
frequency of 50 Hz requires that the system balance of feed-in, off- (iii) Energy cost management (ECM):
take and losses are balanced at any time or that it will be balanced The benefit area is similar to arbitrage at power exchange
in case of any deviations in a short period of time [109]. An markets. In this case not wholesale prices but individual end-user
increase or decrease in net output of BSS can ensure a real-time tariffs are relevant. The BSS can avoid high price energy purchases
system balance [110]. Since 2001, the German TSOs procure their during peak demand hours for residential and commercial/indus-
needs for different control reserves (primary, secondary and trial users [116]. Since 2010, according to Section 40(5) EnWG
tertiary control reserve) on an open, transparent and non-discri- energy suppliers are obliged to offer load-variable and daytime
minatory market. The main differences between the three control dependent tariffs. The tariff-structure and -spreads depend mainly
reserve forms are the tender time and period, the product time- on the respective supplier and individual electrical demand
slice, the award criteria and the remuneration. In addition, positive amounts (e.g. industrial, residential).
and negative SCR and TCR are separately marketed, whereas in the (iv) Reactive power management (RPM):
case of PCR the power increase and decrease must be ensured by a Producers and network operators need to transfer the apparent
single offer, but the forms of control reserve can be provided by power according to the active and reactive power demand of the
various technical units (also known as pooling). end user. Common supply contracts in the industry allow that 50%
(b) Revenues based on bilateral contracts of the active energy can be obtained free of charge as reactive
(i) Voltage support: energy, which corresponds to a cosφ of 0.89 [117]. In case of a
In order to maintain stable network operation, the voltage level higher demand for reactive power an additional fee must be paid,
must be kept in certain ranges. The static voltage support can, which is subject to individual negotiations. This inductive reactive
among others be achieved by a local offset of reactive power [111]. power demand can be covered amongst others by a BSS.
BSS with an inverter and a corresponding power electronic can (v) Demand management:
principally provide reactive power [112]. A compensation of As standard load profiles are applied in the customer segment and
reactive power is exclusively paid on the high and extra high only annual energy consumptions are measured, no tariffs with
voltage level by the respective TSO. On the distribution level the power limits or incentives are available at the moment. This can
requirements are part of the FNN-guidelines but there is no potentially change with the roll out of smart meters. However,
monetary compensation [95]. industrial consumers typically have two price components: expenses
(ii) System restoration: of the peak power demand and expenses for the consumed energy
BSS can be used to energise transmission and distribution lines [103]. Usually, demand management is done by the retraction of
and have the ability to synchronise sub-systems as well as back-up running processes. Therefore, a load-shift via BSS may have (along-
other black start units [105]. In Germany, each of the four TSOs in side with economic aspects) production-related benefits.
cooperation with the DSOs are obliged to have a sufficient capacity (vi) Renewable energy self-consumption (RESC):

1049
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

End-consumers with generation capacity (e.g. photovoltaics) can of PV, for example, are assumed to decrease by 30–50% from 2014 to
increase the amount of self-consumed energy by adding BSS. With 2030 [132]. An even more drastic price decline is foretold for BSS,
the increasing difference between cost of generation and purchase especially for lithium-ion batteries (LIB). The lowest battery cell price
price BSS become more and more attractive to end-consumers. For for utility scale LIB could decrease by 64% from 2014 to 2020 [133].
instance, PV-generation costs and feed-in tariffs have dropped well Although normally only addressed as LIB, there are at least four
below purchase prices from the grid, whereas purchase prices have promising types of LIB suitable energy storage applications with
increased continuously [103]. It is noteworthy that due to the EEG different cell chemistries [134] and price reduction potentials till
amendment from 2014, newly installed systems over 10 kW or 2020 [133]: lithium manganite (39%), lithium nickel cobalt aluminum
10.000 kWh/a are surcharged for own consumption. Overall, the oxide (50%), lithium-iron phosphate (37%) and lithium titanate (25%).
attractiveness of RE self-supply depend highly on electricity fee A more conservative meta-study conducted by Nykvist et al. indicates
regulations. that the costs of LIB for battery electric vehicles could fall below 150
(vii) Grid expansion relief: USD/kWh by 2025, and therefore decrease by more than 50% [135].
Due to the growing energy demand, decoupled supply and The lowest battery cell price for utility scale flow batteries is predicted
demand regions, as well the fluctuating nature of most renewable to decrease by 48% until 2020, making them the second most
energy generation, further investment in new lines, transformers and interesting battery type concerning the price reduction potential [133].
substations may become necessary [118]. According to the usual load The liberalisation of the energy market since the 1990s has not lead,
characteristics, the available transmission capacity limits only the as theoretically predicted, to a decline of the electricity price for
maximum transmittable power, but not the energy [119]. BSS can household consumers due to more competition, but to an increase in
help defer or avoid grid expansions by storing energy. Nevertheless, all 27 member countries of the EU-27, except Finland, between 1998
BSS in general are more cost intensive and the current incentive and 2008 [136]. As electricity prices are much harder to predict than,
regulation (ARgeV) does not consider alternative and perhaps more for example, the LCOE of PV a large variety of methods have been
expansive infrastructure investments. applied over the past 15 years [137], indicating that the electricity price
for households will further rise all over Europe [136]. Keeping in mind
According to a German market analysis based on data from 2013 the big uncertainty of predicting these prices the electricity retail price
the benefits can be grouped in accordance to their market potential (see in Germany is likely to increase until 2030 according to a technical
Table 7). The market potential consists of three core aspects: con- report commissioned by the Federal Ministry for Economic Affairs and
ceivable revenue, applicability for BSS and a favourable legal frame- Energy [138].
work. Only a low potential for BSS benefits lies in grid expansion relief, In countries with lower LCOE of PV compare to Germany like Spain
voltage support and system restoration; redispatch, demand manage- for instance, self-consumption systems might have a positive NPV, but
ment and reactive power management hold a medium benefit potential. a possible back-toll fee could turns a profitable system to a negative
A high market potential is given by energy trading at the day-ahead and NPV [139]. Therefore a favourable legislative framework, as it is the
intra-day market, frequency support, un-interruptible power supply, case in Germany, is mandatory for this business case. By analysing the
balancing group management, energy cost management and renewable Italian market, one can deduce which size is more profitable in a post
energy self consumption. The highest revenue potential for the market feed-in market. It can be concluded that small residential PV systems
based applications lies in the primary control reserve market whereas have higher net present values than bigger systems, as the economy of
the highest cost reduction potential can be seen in maximising the self scale does not compensate the benefits of smaller systems [140].
consumption using renewable energies, especially for households. Therefore, the trend of installing PV systems in LV grids in Germany
Therefore, many BSS projects, especially in Germany, but also world- is likely to continue. PV systems in southern Germany reached PV grid
wide focus on these two applications [120]. An up-to-date world-wide parity in 2012 [2]. With only a PV-system to match the demand, the
database on energy storage systems and their applications is main- achievable self-consumption rates are limited, and can only be
tained by the US Department of Energy [121], which confirms that increased by demand side management (DSM) and BSS come into
these two applications are the most common. Ergo, the focus of this play. It is shown by [141] that BSS have a higher potential to increase
work lies on operating strategies for the maximisation of self con- self-consumption than DSM [141]. Consequently, self-consumption
sumption (Section 5.2) and primary frequency control (Section 5.3). increase is mainly realised with residential energy storages (RES), as
Another approach is to combine complementary business models, this this business case became profitable in 2013 in Germany [142]. As
may increase the profit compared to a single revenue stream [122]. described before, the benefit in 2013 results from the PV LCOE, which
are currently between 9.8 and 14.2 EURct./kWh in Germany [143],
5.2. Detailed overview of operating strategies for self-consumption and the electricity costs for households, which amount to 28.9 ct./kWh
[144]. It is noteworthy that due to the EEG amendment from 2014
With the rise of DG the idea of the prosumer (entities that consume newly installed systems over 10 kW or 10.000 kWh/a are surcharged
and produce), first mentioned in 1980 [123], became more popular. for own consumption (currently with 6.2 EURct./kWh). Therefore, the
The main motivation to become an electrical prosumer as defined in theoretically achievable profit margin lies between 8.5 and
[124], is that self-consumption of locally generated electricity, as 19.1 EURct./kWh. This led to an installation of more than 4600
defined in [125], is more profitable than drawing it from alternative residential storage systems for self consumption in Germany until
supplies. This is the case if the levelised costs of electricity (LCOE) of June 2015 [145].
the DG can compete with the cost to draw electricity from the power In the industry segment the PV generation costs are generally
grid (electricity retail price). A comprehensive manual to calculate the 2 EURct./kWh lower than in household applications because of the
LCOE for renewable energies was first presented by [126] and has larger systems sizes and lie between 7.8 and 14.2 EURct./kWh [143].
further been discussed by [127,128] and [129]. To incorporate the cost The power purchase costs for large customers with a consumption of
of storage [130] proposed to calculate the levelised cost of stored 100 GWh/a range between 4.1 and 15.6 EURct./kWh. Thus, the
energy. theoretical realisable value range (considering the EEG surcharge) is
A comprehensive overview on grid parity world-wide is given by 0–5 ct./kWh.
[131]. It is shown that Europe was the first main market world-wide But could it be economically feasible to pool the prosumer and
where grid parity was achieved in 2010. It is quite likely that the instead of having a BSS and PV-system in every household share and
market volumes for self-consumption business cases will grow in the scale them up? [146] showed that the pooling of prosumer generators
future as the trend of falling LCOE of DG and BSS continues. The LCOE and loads has been beneficial in all calculated scenarios in the UK

1050
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Table 7
Potential benefit estimations for the German electricity market in 2014.

Application Benefit potential Notes

Market based revenues day ahead market 0.00–51.29 EUR/MWh [203] based on mean hourly rates
intra-day market 0.00–69.10 EUR/MWh [203]
PCR 17.60–20.01 EUR/MWh [204]
SCR(pos.) 7.87–11.91 EUR/MWh [204] min.=average power price;
SCR (neg.) 11.83–53.17 EUR/MWh [204] max.=average marginal power price;
TCR (pos.) 0.95–1.58 EUR/MWh [204] potential for SCR and TCR higher
TCR (neg.) 5.72–8.63 EUR/MWh [204] because energy price not included
Bilateral contracts voltage support 0.60–8.70 EUR/Mvarh [205–207] based on available TSOs-price sheets
system restoration 6.85 EUR/MWh [208] high uncertainty, based on US-data
based on marginal cost of conventional
power plants (=cost-based redispatch)
redispatch 9.72–47.54 EUR/MWh [209] min.=marginal costs nuclear;
max.=marginal costs natural gas

Cost reduction UPS 12.72–27.72 EUR/MW/h [210] high uncertainty, based on US-data
BGM: reBAP (pos.) 0.01–43.05 EUR/MWh [16] max.=average volume-weighted
BGM: reBAP (neg.) 0.01–9.39 EUR/MWh [16] reBAP prices; potential ascending
exemplary cost analysis of the “big four”
ECM (households) 9.00–98.00 EUR/MWh (E.on, RWE, Vattenfall and EnBW);
[211–214] difference between high and low tariffs
based on day-ahead market data:
ECM (industry) 10.7 EUR/MWh [203] average price block-contracts, peak load
(hours: 09–20) and offpeak (hours: 21–08)
RPM 13.00 EUR/Mvarh [117] based on capacitor bank prices by [117]
15.00 EUR/MWh based on TSOs power prices on the high
demand management [206,215–217] voltage level <2, 500 h/a
85.00–191.00 EUR/MWh residential PV-system costs [143] and
RE SC (hh) [143,218] av. electricity costs for households [218]
0.00–50.00 EUR/MWh Based on LCOE of large scale PV [143]
RE SC (ind.) [143,144] and el. price for large consumers [144]
grid expansion costs: costs based on the
0.10–0.20 EUR/MWh “Bundesländerszenario” [17]; grid asset
grid expansion relief [17,219,220] lifespan 40 years; consumption in the
distribution network 300 TWh

compared to a single prosumer. This is due to the combination of PV CES do not have to limit the rated power of the DG PrDG. For the
systems, wind turbines and loads. By doing this the self-consumption following graphs it is assumed that the CES is connected to the low
level could be raised up to 17,5%, wherefore the economics in case of voltage and the yearly energy consumption is equal to the energy
grid parity improve significantly. However, BSS were not considered in production of the DG in the LV grid. It is assumed that all DG are PV
this study. Large scale or pooled BSS that apply a self-consumption systems. As suggested in [152] the ratio between capacity and the rated
maximisation can be addressed as community electricity storage (CES), power of the PV system is 1:1. The implementation of the different
as defined in [147,148]. A more detailed definition of CES is given in operating strategies of the German CES projects listed in Table 8 are
[149]. Parra et al. [150] conducted a study in which the LCOE of single sorted in the four categories and described briefly.
households in the UK using PV residential storage systems and using a
CES instead were compared. It has been shown, that the LCOE could be (a) Direct loading
lowered by 37% for a 10-household community and 66% for a 60- The generated energy is directly stored in the BSS if the
household community. In Germany, CES, diverging from the definition residual power Pres of load and generation is positive. This simple
in [148] cannot be operated or owned by the DSO using the CES to strategy maximises the self consumption rate as it ensures that the
participate in the energy market because of the unbundling. The CES BSS is loaded as soon as possible. Drawback of this strategy are the
has to be owned and operated by a citizen cooperative or an external steep gradients depicted in Fig. 7 and that, depending on the
storage operator, for example. In Germany, no similar calculations battery capacity, an excessive feed-in to the grid might occur
considering the potential of lowering the LCOE have been conducted, during peak irradiation around noon, if there is no PV power
but [151] showed that by applying CES the losses caused by the grid- limitation on the power of the PV systems.
compatible storage operation can be lowered by 50% on average A grid compatible operating strategy using direct loading to
compared to RES. With the existing legal framework the business ensure a maximal renewable energy self-consumption rate
models of residential storages and CES cannot be directly compared (RESCR) is used by [153] and [154]. In [153] the BSS is placed
because of the additional burden of extra fees and taxes for CES. in the LV side of a micro grid with DG, which is connected to the
Nevertheless, the studies mentioned before seem to indicate that CES public grid via one MV/ LV transformer. The charging and
have some advantages over residential storages. discharging of the battery is calculated in 1-h steps, from measured
The operation strategies however, can be transferred and classified and synthesised time series. The main differences in the CES
into the following four categories [95,98]: direct loading, schedule project of [154] are that the generation and load of every
mode, peak shaving and based on a prognosis. A more detailed participating prosumer is measured every 5–7 s and that the BSS
description and quantitative comparison of the control strategies for is not necessarily placed at the same location as the DG and
residential systems can be found in [20]. Although being very similar, consumers. The idea of this project is that every participant may
the control strategies for CES are different, as the incentive programme use a part of the battery that is virtually partitioned to increase the
introduced by the German government only supports storage systems individual RESCR.
for grid connected PV systems up to 30 kW [98]. As a consequence, (b) Schedule

1051
 Table 8
Overview of recent large scale BSS projects to maximise self-consumption and peak shaving in Germany.

Project name LCOE; NPV; profit/a Generator (s) Battery Load (s) RESCR Operating strategy/ Ref.
M. Resch et al.

comment

MSG EUREF LCOE=0.52 EUR/kWh PPV = 91 kWp ; LIB: capacity=78 kWh Total yearly energy demand: approx. 64% direct loading/ all values [153]
Pwind = 330 kW PbB: capacity=90 kWh 400 MWh/a for office buildings for scenario 03
supercap: capacity=3 kWh
Strombank Positive NPV, in case of no FIT PPV = 64 kWp LIB: Pc ./ disc . = 100 kW 14 households 4 industrial facilities 30–60% direct loading; prognosis [154,221,222]
and if no taxes and tariffs apply PCHP = 16.5 kW capacity=100 kWh based dis-/charging is
planned
IRENE no profit compared to traditional PPV = 20 MWp ; LIB: Pmax = 5.6 MW(2022) 70% (2022) peak shaving [160,14]
grid expansion Pwind = 23.5 MW (2022) Pc ./ disc . = 70 kW
capacity=162 kWh
Fechheim N/A PPV = 90.5 kWp LIB: 16 households Pmax = 30 kW 100% peak shaving [162]
Pc ./ disc . = 45 kW
capacity=230 kWh
Smart Operator N/A PPV = 60 kWp PbB: 110 households N/A peak shaving [163,223,165]
Pc ./ disc . = 30 kW
capacity=150 kWh
SmartRegion Profit: 150 kEUR/a (2013) PPV = 772 kW ; LIB: 20 households Pmax = 487 kW 93% prognosis based dis-/ [93,71]
Pellworm Pwind = 330 kW Pcharge = 560 kW charging; using external
Pdisc. = 1000 kW forecasts
capacity=560 kWh
VRFB: Pc ./ disc . = 200 kW
capacity=1.6 MWh
EEBatt N/A PPV = 300 kWp LIB: Pc ./ disc . = 200 kW 50 households 25% prognosis based dis-/ [224,151,169]
capacity=200 kWh charging; using
persistence forecasts

1052
Smart Grid Solar N/A PPV = 287 kWp PbB: Pc ./ disc . = 72 kW 16 households N/A prognosis based dis-/ [225,179]
(Epplas-Hof) Pdisc. = 72 kW charging
capacity=330 kWh
Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

In this strategy, the time to charge the battery will be shifted to proposed in the next strategy.
a typical time with high radiation. The schedule mode with
constant charging power is depicted in Fig. 8 showing a more (a) Prognosis based strategy
favourable behaviour from grid perspective because feed-in peaks This strategy uses load and weather forecast data to adjust the
as in the direct loading strategy are prevented. charging power and feed-in power to get a fully charged battery at
Nevertheless, the self-consumption rate might be reduced, as in the end of the day and/or avoid over-voltage and asset over-
days with lower radiation the BSS might not be fully loaded. loading (Fig. 10).
Several strategies have been proposed for this purpose. The main
differences are that [155] and [156] propose a starting point A control loop within the day corrects the deviation from the
around noon and charge the battery with full power whereas forecast data. This strategy reaches the highest self consumption rate
others, for example [157], suggest a constant charging power over after the direct loading strategy while still being grid supportive, this is
a larger period. Currently there is no CES project in Germany due to the lower curtailment losses compared to other strategies
known to the authors using this strategy. [20,125]. The main differences of this strategy are the forecast
(c) Peak-shaving (load levelling) techniques. Principally, the previously published studies can be divided
The main objective of the peak shaving strategy (Fig. 9) is to into four classes:
avoid over-voltage and equipment over-loading issues by limiting
the power at the PCC and using the remaining residual power to (i) Studies using a perfect forecast [152,166].
charge the battery [158,152,159]. (ii) Studies using synthetic forecasts (modified measured time series)
[167–169].
The limitation of power at the PCC should be based on the voltage (iii) Studies based on external weather-based forecast from meteor-
at the PCC, the power range of the battery, and the PV penetration of ological services [170,171,161,172].
the grid [158]. The main objective of this strategy is to not surpass a (iv) Studies that base their forecast on a persistence method based on
certain level of Pres / PrDG at the PCC of the BSS. There are mainly three values measured by the PV-system [167,125,173].
possibilities to achieve this aim:
Obviously, no prediction errors apply to a perfect forecast. The only
(i) The battery is sized for the worst case, e.g. the day with the highest difference is the time resolution, which in the case of [152] is 1 min and
irradiation and no load, as in [160]. in the case of [166] is 15 min.
(ii) The power of the DG is curtailed in case of a full battery, as One proposition for modelling synthetic forecast which has been
depicted in Fig. 9 and described in [159]. presented by [168] and also used by [169], uses the Spherical
(iii) Instead of curtailing the DG an additional load is used to reduce Harmonic Discrete Ordinate Method [174]. In this model measured
the residual load by using, for example, power-to-head [161]. data is used to generate the global solar irradiation at ground level for
the next days. By forecasting the weather data a minute-based PV
This grid supportive operating strategy is applied to large scale BSS power is calculated taking into account the orientation and angle of the
by [14,160,162] and [163]. The focus of IRENE Project lies on grid power plant. An error analysis of the model has shown that the average
expansion relief. Therefore, one or several BSS are dimensioned and error (rRMSE) of the weather forecast for the next day is 32.5% for one
placed strategically in the LV to mitigate the total feeder RPF to 70% of site. This is very close to the accuracy of approximately 30% of current
the cumulated PrDG of the respective feeder in which the BSS are numerical weather prediction models for Central Europe [175]. The
installed. Additionally to this active power control, a reactive power value increases for a longer forecast horizon. Instead of a physical
control is implemented. The calculation of the set-points of P and Q are model, [167] uses a noise sequence to fabricate a forecast based upon
calculated externally and not by the BSS itself [160]. the hourly average of the measured data. This results in an hourly
Similar to the aforementioned project the BSS in Fechheim limits forecast for the next day with an rRMSE of 30%.
the power to 40% of the cumulated PrDG of feeder in which the BSS is Several studies use external weather forecasts and calculate the AC
allocated with an active power control and uses a reactive power power profile of the PV system according to predicted irradiance
control to reduce the voltage in the case of a fully loaded storage [162]. instead of synthesising the forecast data. A simple forecast method in
The aim of the SmartOperator project is to minimise voltage which the historical data of the solar irradiance and the predicted
deviations and line utilisation. The BSS is dimensioned to enable a weather conditions (sunny, cloudy, rainy) are used to calculate the PV
peak shaving of 50% for a period of 5 h [163] based on initial studies by profile in 1 h steps is presented by [170]. Also on an hourly basis, [171]
[164]. A learning algorithm is used to calculate the forecast of predicts the PV power output for different region sizes in Germany
generation and load as well as future grid states, based on real time based on forecasts for up to three days ahead that are provided by the
data of voltage and current [165]. This forecast is used to calculate the European Centre for Medium-Range Weather Forecasts (ECMWF). For
active and reactive power flows of the BSS to ensure peak shaving of the a single site and day ahead forecast an rRMSE of 36% could be
PV systems and that the voltage values of the grid nodes stay within quantified. As the rRMSE decreases as the examined area rises for the
given thresholds. whole of Germany the accuracy of the rRMSE is 13%. Another study
The advantage of peak shaving is that critical voltages might be uses the irradiance forecast based on the Weather Research and
avoided by limiting the feed-in power. The voltage can be further Forecasting (WRF) Model [176] and evaluated the deviation of the
reduced by absorbing reactive power. From the point of view of self measured irradiance values of a pyranometer (5–8%) and the PV power
consumption maximisation, a problem is raised during cloudy or foggy output (3–5%) on a 15 min base for a PV plant in Italy [172]. Historical
days, when there is not enough radiation to charge the battery. forecast data of irradiance and temperature in 1-h steps from Meteotest
Consequently, the self-consumption rate will be reduced. On the other [177] has been used by [161] to calculate the PV output power and it is
hand, during high irradiance days, the power curtailment is high as can shown that the RESCR decreases by 15% if forecast errors are taken
be seen in Fig. 9 for case (ii). For case (i) and (iii), however the into account instead of assuming a perfect forecast.
additional investment costs have to be considered critically. This Another approach is to use persistence weather forecast. The
applies in particular for case (i) in a distribution grid with many wind forecasting method is based on extrapolating the current or recent
generators as in this case the energy to power ratio of the BSS needs to PV power plant output taking into account the changing of the sun
be higher as for PV systems [164]. To avoid these losses or additional angle. Since the persistence is based on stochastic learning technique
invest, an optimisation of the power flow based on a prognosis is from historical pattern, the accuracy highly depends on the forecast

1053
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Fig. 7. Operating strategy direct loading (generator perspective).

Fig. 9. Operating strategy peak-shaving (generator perspective).

where the optimisation for the day is conducted only once and the
correction due to forecast errors is done by comparing the forecasted
SOC with the measured SOC. Thus, inaccuracies may occur as the SOC
cannot be measured directly, but is a calculated value from the battery
management system.
The differences of the forecast and operation strategy lead to similar
curtailment losses as in [125], but the self-sufficiency is higher with the
adaptive forecast approach of [173]. It could be shown, that the
adaptive forecast shows advantages over a persistence forecast with a
fixed horizon. As the control algorithm based on autonomous persis-
tence forecasts reaches similar curtailment losses as the one based on
external forecasts [173], these forecasts seem to be preferable as they
Fig. 8. Operating strategy schedule mode (generator perspective). need no additional hardware and are independent of the additional cost
of external forecasts or meteorological services. Nowadays, up to one
horizon due to the change of cloudiness [178]. The forecast method is third of the installed residential PV storage systems in Germany are
suitable for minute based forecasts for one location. For simulation capable of applying a prognosis based charging algorithm [95,145].
purposes, an autonomy forecasting using a learning algorithm is more A prognosis based operating strategy is applied to large scale BSS
preferable compared to the one that depends on the global weather by [93,71,151,169] and [179].
data. The differences of the different persistence forecasts arise in the In the project SmartRegion Pellworm different business models
algorithms used to predict the load and PV output and the values that have been tested and affect the operating strategy. The scenario which
are used to correct the intra-day deviation from the forecasted values. maximises the RESCR is called “Sustainable Regional Load Supply”.
As described before, [167] uses a synthesised forecast data with a noise The active power flow of the BSS is an output of an optimisation to
and a learning algorithm based on historical data to adapt the charging maximise the profit for the different business models [93] and is
algorithm to the PV output and load within the day. Also [169] uses a combined with an OPF simulation to incorporate grid restrictions to
synthesised PV forecast; concerning the load an easier method is calculate the reactive power flow [71]. The prognosis is carried out
proposed by predicting it based on the load profile of the past five days. using a perfect foresight based on measured time-series for load and
In this method, the day is divided into three periods: midnight to generation and historic market data. This central approach ensures a
sunrise, sunrise to sunset and sunset to midnight. Using the arithmetic grid supportive behaviour of the BSS.
means of the past five days, the load profiles of each period define the The EEbat project combines the aim to relieve the grid and
load for the next two days. maximise self-consumption. The grid relief is achieved, by applying
Fully autonomous persistence, which is not dependent on an the peak shaving strategy using active power control and curtailing the
external or synthesised forecast is presented by [125] and [173]. In RPF to 50% of the cumulated PV installed in the LV grid. A persistence
[125] a method is used that assumes a load profile for the predicted forecast for load and generation is used to adjust the charging power
weekday identical to the load profile of the weekday from the previous and ensure a maximum self-consumption [169]. Furthermore, the
week and predicts the PV for the next day based on the day before. To
correct prediction errors within the day a proportional plus integral
controller (PI-controller) adjusts the feed-in limit by constantly
comparing the difference between target and actual SOC. The load
prediction in [173] is the same as before-mentioned. The study also
shows that the forecast of the PV output has a stronger impact on the
curtailment losses and self-sufficiency rate than the load forecast
compared to a perfect load forecast. Therefore, an elaborated method
for the PV persistence forecast is presented: First a bell-shaped profile
based on the last ten days is calculated. To achieve a higher accuracy a
moving horizon is introduced that combines the PV data from the last
4.5 h with the bell-shaped profile. For the intra-day correction the feed-
in limit is adapted dynamically every 15 min by running an optimisa-
tion with 15 min of forecast resolution and 15 h of optimisation
horizon, if the measured values (residual load and battery charge
power) differ from the predicted. This differs from [125] and [169] Fig. 10. Prognosis based operating strategy (generator perspective).

1054
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

differences between using standard load profiles (SLP) [180] and storage capacity compared to other storage technologies such as pump
realistic load profiles as inputs for the operation strategy are quantified. storage systems [111], this storage capacity is fully sufficient (when
For this calculation the RESCR and the financial benefit of a CES made sure that the state of charge (SOC) of the battery is kept at an
allocated in a LV grid consisting of 50 households with PV generation optimal level during operation (see below)) to bypass the time until
and loads are compared. It is shown, that SLP are sufficiently accurate primary control reserve is relieved by secondary control reserve (see
to be used as input for this operation strategy. Reactive power control is Fig. 12) [186]. The need for a relatively low storage capacity of course
not considered in this study. also has the benefit of reducing investment costs and hence has a
Another charging strategy is implemented in the Smart Grid Solar positive effect on the profitability of the battery system.
project. To predict the PV generation, a short term weather prediction From a financial point of view, however, there are further points
based on sky-images instead of measured electrical values is imple- that make the provision of PCR the most attractive business case for
mented. The charging algorithm is rather simple as the battery is large scale batteries nowadays [40]. As shown in Section 5.1, the main
charged with a constant charging power in case the residual power reason for this is, that under the actual economical and legal frame-
exceeds a given limit. Another control strategy developed in the same work, the weekly income is the highest when compared to other
project is based on the measured voltage at the PCC which is kept business cases. Because of this, it is foreseen that already existing
within a given range by charging or discharging the battery [179]. PCR battery projects will turn out as being profitable in the near future
The impact of different operating strategies on distribution grid [40,186]. Another argument making the provision of PCR with large
planning is discussed in Section 5.4. scale BSS very attractive from an investor's point of view is the already
existing PCR market with its clear rules. This on the one hand reduces
5.3. Detailed overview of operating strategies for primary frequency the risk for future income expectancies and on the other hand lowers
control marketing expenses.
The mentioned technical as well as economic reasons for providing
Due to the fact that there is only very limited possibility of storing PCR with large scale batteries have led to an increment of existing as
electric energy in the electrical system nowadays, a constant equili- well as planned primary frequency control battery projects in Germany
brium between active power generation and consumption must be over the last years. A chronological overview of recent large scale BSS
maintained. An indicator for the deviation in this balance is the system projects for primary frequency control in Germany are listed in
frequency, since it is a measure for the rotation speed of the Table 10. The first battery providing PCR within the European grid
synchronised generators. An increment of the total load will decrease was a NAS battery. This battery was integrated into the German
the speed of the generators and hence lower the system frequency. A network in the year 2012 by the Younicos AG. As can be noticed, since
decrease of the demand on the other hand leads to an increase of the then the installed power of the battery systems has been steadily
system frequency. [181]. increasing. Furthermore, it can be derived from Table 10 that almost all
Since frequency deviations can not only damage electronic devices projects apply Li-Ion technology. One of the main reasons for this are
connected to the grid but also endanger the stability of the whole the rapidly falling costs for Li-Ion batteries over the last years
electrical network, the German transmission system operators (TSOs) [186,187]. Besides this, Li-Ion batteries have also one of the highest
are legally obliged to maintain the system frequency within the strict roundtrip efficiencies in comparison to other battery technologies, a
limits of 50 Hz ± 1% (see also Section 2) [23,26]. In order to achieve very high energy density, high lifetime expectancy as well as a very
this goal, a certain level of active power reserve is required to re- favourable power to energy ratio for providing PCR [188,189]. This
establish the equilibrium between demand and generation in case of means that a high installed power does not lead to an unnecessarily
unbalances (this can be unbalances between instantaneous power high storage capacity. Nonetheless, flow batteries in primary reserve
consumption and generation, but also major power disturbances in applications have also been discussed in literature [190]. The same
the grid) [181]. author claims that short response times as well as the ability of some
The “Operational Handbook” of the ENTSOE (European Network systems of being overloaded give BSS an advantage over conventional
of Transmission System Operators for Electricity), which sets general facilities. As more and more private companies plan PFC projects
rules and technical recommendations regarding reserve power levels without federal funding one can deduce that this business case seems
and their associated control performance, defines three different promising from their point of view and is technically mature. Still, the
reserve levels: primary, secondary and tertiary control reserve pre-qualification that allows the facility to operate at the PFC market is
[182,183]. According to the Grid Code of the German TSOs these the bottle neck at the moment, as most of the commissioned projects
reserve levels are also valid in Germany [182]. The primary control did not pass the pre-qualification yet. Another trend is the increase of
reserve (PCR) is automatically activated within a few seconds after the system size of large scale BSS as it can be seen for the most recent
detecting a frequency deviation according to the curve depicted in systems under construction in 2016.
(Fig. 11). Whether the number of grid connected large scale battery systems
The main goals of the secondary control reserve (SCR) are to will continue to rise in the future depends to a great extent on prices
restore the rated frequency of the system, to release primary reserves decline for batteries and the future development of the remuneration
and to restore active power interchanges between control areas to their for primary control reserve. Since the request for batteries has steadily
set points. The tertiary control reserve (TCR) aims to replace the been increasing over the past years, battery costs are generally expected
secondary reserve, manage eventual congestions and bring back the to fall in the future [185,191,192,133]. For a more detailed cost
frequency to its rated value if secondary reserves are not sufficient [27]. prognosis please refer to Section 5.2) The future development of the
In Fig. 12 the interaction as well as the starting and deployment remuneration for PCR, however, is relatively unclear since it depends
times for the three reserve levels according to the guidelines of the on many factors that are barely predictable. These are for example the
German Grid Code is shown [27]. number of players in the PCR market and the future request for
In Germany, large scale battery storage systems are almost exclu- primary reserves. In [193] and [194] it is estimated that the future
sively used to provide PCR. There are several technical as well as request for primary reserves will rise due to an expected increase of the
economical reasons for this. From a technical point of view batteries share of fluctuating renewable energy sources along with their low
perfectly suit the operational requirements for providing PCR since predictability of electricity production. In [40] and [195] on the other
they are able to deliver the requested power very accurately within a hand it is estimated that the request for primary reserves will stay more
time frame of less than one second with a very high reliability or less constant in the future. This is explained by the fact that the
[184,185]. Although large scale batteries usually have a very limited demand for PCR in Europe is actually determined on the basis of the

1055
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Table 9
Key parameters for the provision of primary control reserve [27,226,227].

Max. frequency response insensitivity ± 10 mHz

Full activation frequency deviation ± 200 mHz
Full activation time 30 s
Tendering period 1 week
Min. bid size ±1 MW
Time availability 100 %

projects uneconomical. Because of this, the German TSOs have defined

“degrees of freedom”, which give battery operators the chance to
readjust the SOC of the storage system during operation [197]. As a
consequence, the required storage capacity is reduced, since the SOC
Fig. 11. Relation between frequency deviation and provided primary control reserve. can be kept at a level, where it is ensured that the battery is able to
provide the requested balancing power until primary control reserve is
relieved by the secondary control reserve Fig. 12. For this case a power
to energy ratio of one (e.g. 1 MWh/MW) is fully sufficient [186,187].
According to [184] and [187] the optimal SOC for batteries
providing primary control reserve lies around fifty percent. The reason
for this is that the network frequency generally fluctuates more or less
normally distributed around the nominal value of 50 Hz [103].
Therefore, approximately the same amount of balancing power has to
be provided in positive (unload) as well as negative (load) direction.
Fig. 12. Starting and deployment times of primary (PCR), secondary (SCR) and tertiary
control reserve (TCR). Due to the losses of the storage system, however, the SOC tends to fall
in the long run. Hence, it is advisable to keep the SOC slightly above
fifty percent [184]. The TSOs in total defined six degrees of freedom for
simultaneous loss of the two largest power plants within the European SOC adjustments. They can be found in [197]. The main difference
grid, which is not expected to change significantly in the future. A between them is that some generate extra costs for the battery operator
comprehensive study on how the rise of variable renewable energies and some do not. Those degrees of freedom that do not generate costs
and the reserve market interacted in Germany in the past years is given can be applied as often as required. Those that do generate costs on the
by [115]. Hirth and Ziegenhagen [115] try to explain the possible other hand should be applied as seldom as possible. In this case the
reasons of the reduction of the balancing reserves and costs and the decision whether to use the degree of freedom or not becomes more
simultaneous increase of installed wind and solar power. One of the complex and should be determined on the basis of a cost benefit
major findings is that the wind and solar power forecast errors might calculation. All six degrees of freedom listed in [197] are briefly
not be the most prominent driver for the balancing reserve require- described hereafter (italic letters): As can be seen in Fig. 13 the
ment, but that other factors like the design of the control market might “optional overfulfillment” gives the battery operator the chance to
be more important. Due to all these uncertainties, the prediction of the provide 20% more balancing power than required, if it is useful for an
price development for PCR is hardly possible and expert opinions adjustment of the SOC.
strongly differ in this point [185,195,115]. The degree of freedom “dead-band” makes it possible to readjust
Another important factor that can have a big influence on the the battery SOC by using the dead-band (Fig. 14). One condition for the
development of the number of large scale batteries in the German grid application of this degree of freedom is that the behaviour of the
are future adjustments of the participation conditions for the PCR battery must always support the stability of the electrical network,
market. On their basis it is not only decided who is able to enter the meaning that, for example the battery is not allowed to charge when
market and who is not, but they also set the operational framework for positive primary control reserve (unload) is required.
PCR providers. On the other hand, this can have a big influence on the One degree of freedom that has to be remunerated when applied is
economics of PCR projects. For example, if the required storage the option to charge or discharge the battery with “schedule transac-
capacity of PCR batteries has to be increased, as it is currently tions”. In this case the SOC can be optimised by purchasing or selling
discussed [185,196], it would have a negative impact on the economics energy at the energy market (stock market or over the counter
of those projects. transactions). Of course, when using this degree of freedom the battery
The guidelines for entering the PCR market are defined by the operator has to make sure that the sum of battery output and
TSOs, since they are legally obliged to ensure that all technical purchased / sold energy corresponds exactly to the requested value
standards for operating the electrical network are safely fulfilled [23]. by the TSO at any point in time. An exemplary behaviour of the battery
The actual key parameters for the provision of PCR are summarised in during a schedule transaction is shown in Figs. 15 and 16. In this case
Table 9. Furthermore, according to the German Grid Code all the SOC of the battery is in its lower half at 8:00 o'clock. Since the
prospective providers of PCR have to complete a pre-qualification battery has to keep continuously unloading due to low grid frequencies,
procedure to demonstrate their ability to meet the requirements in this a schedule transaction is carried out between 9:00 and 9:15 o'clock. As
respect [27]. can be noticed, this prevents the SOC from reaching a critical value,
As can be seen in Table 9, the primary control reserve has to be since the battery is loaded instead of unloaded in this time window (see
provided for a tendering period of one week with an availability of one Fig. 16).
hundred percent. For battery storages this would mean that they would Similarly to the just described degree of freedom it is possible to
have to be dimensioned for the case that the full offered power is “load or unload the battery with another technical unit”. One
requested continuously during a whole week. The dimensioning for this condition for doing this is that all entities involved in the re- or
unrealistic worst case scenario, however, would make all battery discharging process must belong to the same balancing group.

1056
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Table 10
Overview of recent large scale BSS projects for primary frequency control in Germany, based on [121] and contact with the BSS owners.

Project name BSS type Rated power Duration at Pr Commissioning date Funding Pre-qualified Lifetime [a]
Pr [kW] [HH: MM] source

WEMAG Younicos Battery Parka LIB 5000 1:00 16/09/2014 federal/ yes 20
private
Younicos and Vattenfall Project: Sodium Sulfur / NaSB/LIB 1000/ 200 6:00/ 1:00 01/12/2012 federal/ yes 20
Lithium Ion private
M5BAT (Modular Multi-Megawatt Multi- PbB/LIB 5000 1:00 mid 2016 federal/ no 2
Technology Medium-Voltage Battery Storage) private
Feldheim Regional Regulating Power Station LIB 10,000 1:00 21/09/2015 federal/ no 10+x
(RRKW) private
Bosch Braderup ES Facility LIB/ VRFB 2000/ 325 1:00/ 3:50 11/07/2014 // 15/09/ private no 15
2014
1.3 MW Battery in Alt Daber advanced PbB 1300 0:40 10/2014 federal/ yes 15
private
Bosch Second Life Batteries LIB 2000 1:00 Q3 2016 private no 10
REDMONDIS Electrorecycling Plant LIB 13,000 1:00 mid 2016 private no 10+x
3 MW Battery Storage - Dörverden, Germany - LIB 3000 N/A 15/12/2015 N/A N/A N/A
Statkraft
LESSY LIB 1000 0:42 01/02/2014 federal yes N/A
90 MW Energy Storage - STEAG GmbH LIB 6 x 15,000 1:30 mid 2016 to early 2017 private no N/A
SmartPowerFlow VRFB 200 2:00 02/09/2015 federal no 1

a
First stand alone battery in Europe, according to [228].

Furthermore, an optimal interaction of the involved units has to be diversity factors for implementing BSS in the grid planning are
demonstrated in advance. listed in Table 11 and result in a neutral behaviour of the BSS. The
Another degree of freedom for batteries consists in the “relocation operation strategy direct loading and schedule as used in the
of the dead-band when grid-time corrections are planned”. When projects Strombank and MSG EUREF (see Table 8) can be men-
required the PCR provider is informed one day in advanced about the tioned as an example.
target frequency for the upcoming day by the TSO. In this way the PCR • grid supportive self consumption
provider is able to prepare the dead-band shifting for the time period of For the HLF the same as for grid compatible BSS applies, as the
the grid-time correction. battery might be fully discharged as well. The difference arises for
According to [181], the maximum deployment time for PCR the RPF. In this case the battery is used to mitigate the reverse
increases linearly with the requested primary control power. Starting power flow caused by DG with peak shaving. The peak shaving
from a value of zero the maximum offered power by a PCR provider threshold can be either fixed or adaptive as in the case of forecast
must be fully activated after 30 s at the latest. However, BSS that are based charging and discharging. For the projects listed in Table 8
able to provide the requested power much faster are allowed to use this that use a forecast based operation mode a peak shaving function-
characteristic as a degree of freedom. This means that battery operators ality is implemented. Nevertheless, the rated power of the BSS might
are allowed to use the whole “permissible operating range” depicted in be higher than the power used to mitigate the power at the PCC,
Fig. 17 to readjust the SOC of their storages. which is the case in the EEBatt project where the energy to power
ratio of the BSS is 1:1. In this case, the diversity factor is the
5.4. Impact of BSS maximising self-consumption and applying PCR quotient of the power used for peak shaving purposes and the rated
on distribution grid planning charging power of the BSS. For example, for the project
SmartOperator (pure peak shaving) the diversity factor is 1, but it
In this section the impact of the operating strategies derived from is <1 in the EEBatt project (forecast based SC). This operating
the business cases of self-consumption maximisation and primary strategy can solve over-voltage (cable) and thermal issues (cable and
control reserve as shown in Sections 5.2 and 5.3 on distribution grid secondary transformer) if the BSS is installed in the same LV feeder
planning are discussed. How BSS can be implemented in traditional as the DG causing them. The thermal load of the primary transfor-
grid planning as presented in Section 2.2 is subject to ongoing research. mer is reduced in any case independently of the allocation of loads,
However, [198] gives some hints by showing that DSO only consider DG and BSS, as the peak of the RPF is mitigated in any case, if a
active power flows, which seems a viable proposition as they are diversity factor of >0 for the BSS is reached.
responsible for the revenue stream for the two business cases and a • system compatible primary control reserve
reactive power control is not yet mandatory for large scale BSS. It can be deduced from Fig. 18, that a BSS providing PFC might
Therefore, in the fist part of this section only the active power flows discharge or charge with its full rated power at any moment.
are evaluated using the worst case approach of traditional grid Depending on the system architecture, some BSS have the capability
planning and the resulting diversity factors for BSS are listed in to be overloaded, as reported in [190] for VRF (100% over-loading),
Table 11. Secondly, the effect of reactive power control on the planning in [184] for LIB (30% over-loading for 15 min), and in [199] (25%)
is discussed briefly as it can be considered independent of the business also for LIB. In a worst-case scenario, the normal operation together
case, given that the power electronics is able to provide a four with the application of the degrees of freedom as described in
quadrants operation. In the last part deficiencies of the traditional Section 5.3 can lead to a diversity factor >1. Depending on the
planning methodology are presented and possible steps to new plan- allocation of the BSS it might reduce the hosting capacity of DG of
ning approaches, as explained in Section 4.1, are discussed. the affected grid as this operation strategy tightens the over-voltage
and over-loading issues. All projects listed in Table 10, except the
• grid compatible self consumption SmartPowerFlow project, where the BSS behaves in a system
The worst case is that the battery is fully loaded for the RPF supportive way fall into this category.
scenario and fully discharged in the HLF-case. The resulting • system supportive primary control reserve

1057
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Fig. 16. Schematic SOC profile for “schedule transactions”.

It can be concluded that the traditional planning method of passive

distribution systems for large scale BSS will lead to over-capacities and
uncertainties concerning the reactive power flows. Therefore, CIGRE
promotes the shift to active distribution systems as defined in CIGRE
Fig. 13. Degree of freedom “optional overfulfillment”.
WG C6.11 [200], which will incorporate DG and BSS in a more active
way than the fit-and-forget approach which is currently used and will
allow to apply new planning approaches more efficiently. This transi-
tion is described in detail by [46]. As discussed in Section 4.1, BSS, as
well as DG and the distribution grids need to be modelled to calculate
time-series and derive suitable probability density functions.
Depending on the application and technology different time-steps need
to be realised in these models [46]. In [201] it is shown that for SC the
operation strategy should be simulated at least in one minute time-
steps to avoid short-term feed-in peaks. For PCR the resolution has to
be even higher and one second time-steps seem appropriate, in order to
incorporate all degrees of freedom described in Section 5.3 properly.
As for the reactive power control current studies focus on two main
directions: a central approach using an AC OPF, such as [71], or an
autonomous voltage control, as for example a Q(V) control [41]. It
seems as if autonomous voltage control strategies are the more
favoured solution at the moment as the technical standard for
connecting BSS and DG in MV and LV are aiming in this direction [31].
The challenges of future investigation lie in modelling BSS to
Fig. 14. Degree of freedom “dead-band”. calculate active and reactive power time series for different applications
in order to apply them for new planning approaches in active
The diversity factor for this operating strategy is the same as for distribution systems.
the gird compatible behaviour, as the active power flows are the
same. The difference here is that a reactive power control is used to 6. Conclusion
solve over-voltage issues.
In this paper, traditional approaches and recent advances in
In the traditional distribution grid planning reactive power control distribution grid planning alongside with alternative possibilities to
is usually not considered and only a fixed cosφ can be taken into traditional grid extension with large scale battery storage systems are
account as only one time-step for the two worst case scenarios is described. In addition the German energy storage market is analysed
calculated. In a grid/ system compatible behaviour cosφ may be set to 0 and the operation strategies of the two most profitable applications,
and in a grid/ system supportive behaviour to the maximum favourable self-consumption maximisation and primary frequency control, are
value from grid perspective. Nevertheless, this issue has not been
analysed systematically yet and may lead to wrong results if the method
of the traditional planning is applied. For an accurate simulation of a
reactive power control, a load flow analysis based on time series has to
be applied.

Fig. 17. Degree of freedom “permissible operating range”.

Fig. 15. Degree of freedom “schedule transactions”.

1058
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

Table 11
Diversity factors for BSS applied for SC and PCR.

HLF RPF

grid compatible SC 0 0
grid supportive SC 0 (0–1)charge
system comp./ supp. PCR (1+x)charge (1+x)disch.

described in detail after an extensive literature review. The main

findings and contributions of the paper are:

• A clear methodology for grid extension measures in distribution

grids has been presented.
• Most of the new approaches for distribution grid planning use
deterministic models and do not consider reliability issues. There is Fig. 18. Statistical requests of PCR power in the UCTE grid [103] (adapted).
a great variety of these models with their respective pros and cons
that have to be considered for the given planning task. Nevertheless,
it is shown that the over-sizing problem remains even for advanced
grid planning methods if worst case scenarios are applied. Furthermore, not all the applications and operating strategies are
Therefore, there is a great need for detailed models to generate mitigating the problems of the DSO that arise with increasing
combined active and reactive power flows of BSS that are market- penetration of DG. Future studies should concentrate on combining a
driven and grid/ system supportive at the same time. profitable and a grid supportive behaviour into one operation strategy,
• An analysis of 20 potential revenue streams for BSS shows that the otherwise the implementation of BSS in distribution grids might lead to
primary control reserve market holds the highest revenue potential further grid extension instead of grid relief.
for market based applications, whereas the highest cost reduction
potential lies in the maximisation of the self-consumption using Acknowledgement
renewable energies, especially for households.
• As suitable options for the maximisation of self consumption the This work was supported by the German Federal Ministry of
operation strategies direct loading, schedule mode, peak shaving, Economics and Technology (BMWi) and the Projektträger Jülich
and prognosis based loading were identified. Additionally, several GmbH (PTJ) within the framework of the project “SmartPowerFlow”
large scale BSS projects in Germany applying those strategies were (FKZ0325523A). The author would also like to thank Mrs. H. Krauth
presented. The prognosis based operation strategy with a peak limit and Mr. A. Penzkofer for the kind collaboration in the proofreading.
restriction seems to be the most promising, as it leads to manageable
curtailment losses, especially if the feed-in limit is reduced in the References
future. Within the forecast based strategies the adaptive forecast
algorithm combines the advantages of autonomy from external [1] Böhme D, Dürrschmidt W, Van Mark M. Erneuerbare Energien in Zahlen, Tech.
forecasts with their accuracy. Nevertheless, due to additional fees rep., Federal Ministry for Economic Affairs and Energy (2014).
and taxes applying for community electricity storages, this business [2] Karakaya E, Hidalgo A, Nuur C. Motivators for adoption of photovoltaic systems at
grid parity a case study from southern Germany. Renew Sustain Energy Rev
case is hard to transfer from residential to large scale storages. 2015;43:1090–8. http://dx.doi.org/10.1016/j.rser.2014.11.077.
Besides a need of revising the existing legal framework in order to [3] Federal Ministry of Economic Affairs and Energy (BMWi), Federal Ministry for the
make CES economically feasible, there is also a need for future Environment Nature Conservation, Building and Nuclear Safety (BMU),
Energiekonzept für eine umweltschonende, zuverlässige und bezahlbare
research regarding this application. Nonetheless, it seems especially Energieversorgung (2010).
interesting as in the near future PV systems, that have reached the [4] Keane A, O'Malley M. Optimal allocation of embedded generation on distribution
end of their 20 year period of feeding into the grid with a fixed feed- networks. IEEE Trans Power Syst 2005;20(3):1640–6. http://dx.doi.org/10.1109/
TPWRS.2005.852115.
in tariff, can be used for this applications with an extreme low LCOE.
•
[5] Ackermann T, Andersson G, Söder L. Distributed generation a definition. Electr
Primary frequency control seems to be the most promising business Power Syst Res 2001;57(3):195–204. http://dx.doi.org/10.1016/S0378-7796(01)
case for BSS in Germany at the moment. Although the net present 00101-8.
[6] Gonzalez C, Ramirez R, Villafafila R, Sumper A, Boix O, Chindris M. Assess the
value is just becoming positive, there is still a great challenge to impact of photovoltaic generation systems on low-voltage network: software
make it profitable. Although the degrees of freedom help to achieve analysis tool development, in: Proceedings of the 2007 9th International
this goal, research is still necessary to determine the different Conference on Electrical Power Quality and Utilisation, IEEE, 2007, pp. 1–6.
〈http://dx.doi.org/10.1109/EPQU.2007.4424183〉.
benefits of these options, especially for VRFB, since most of the
[7] Nykamp S, Molderink A, Hurink JL, Smit GJ. Statistics for PV, wind and biomass
BSS used for primary frequency control are lithium-ion batteries. generators and their impact on distribution grid planning. Energy
• The task to implement BSS in (traditional) distribution grid plan- 2012;45(1):924–32. http://dx.doi.org/10.1016/j.energy.2012.06.067.
[8] von Oehsen A, Saint-Drenan Y-M, Stetz T, Braun M. Vorstudie zur Integration
ning is also subject to ongoing research. At the moment many of the
großer Anteile Photovoltaik in die elektrische Energieversorgung - Ergänzte
studies only consider active power flows and worst case assumptions Fassung vom 29.05.2012, Tech. Rep. November 2011, Frauenhofer IWES, Studie
are applied. If traditional planning methods for passive distribution im Auftrag des BSW - Bundesverband Solarwirtschaft e.V., Kassel (2012).
systems are applied for large scale BSS, over-capacities will probably [9] Mohring A, Michaelis J. Techno-ökonomische Bewertung von Stromspeichern im
Niederspannungsnetz, Tech. rep., Fraunhofer ISI, Karlsruhe (2013).
be the result. In order to evaluate the potential of BSS to behave in a [10] Müller B, Möller C, Gaudchau E, Schneider L, Resch, M, Plessmann G, Breyer C,
grid supportive manner, power flow simulations considering opera- Gajkowski K. Vergleich und Optimierung von (de-)zentral orientierten
tion strategies for active and reactive power control for different Ausbaupfaden zu einer Stromversorgung aus EE in Deutschland, Tech. rep.,
Reiner Lemoine Institut gGmbH, Berlin (2013).
time scales depending on the application have to be conducted. [11] Stetz T, Kraiczy M, Diwold K, Braun M, Bletterie B, Mayr C, Bründlinger R, Noone
B, Bruce A, Macgill I. High Penetration PV in Local Distribution Grids Outcomes
In conclusion, it is worth pointing out that large scale BSS are of the IEA PVPS Task 14 Subtask 2, Tech. Rep. July (2014).
[12] Stetz T, Rekinger M, Theologitis I. Transition from Uni-Directional to Bi-
becoming economically feasible in Germany, however there is a lack of Directional Distribution Grids, Tech. rep. (2014).
planning guidelines for DSO to integrate the BSS in their grid.

1059
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

[13] Kerber G. Aufnahmefähigkeit von Niederspannungsverteilnetzen für die University of Kassel (2014).
Einspeisung aus Photovoltaikkleinanlagen, [Ph.D. thesis], TU München (2011). [42] Agora Energiewende, Stromverteilnetze für die Energiewende, Tech. rep., Berlin
[14] Meyer B, Mueller H, Koeberle R, Fiedeldey M, Hoffman C, Bamberger J. Impact of (2014).
large share of renewable generation on investment costs at the example of aw [43] Union of the Electricity Industry (EURELECTRIC), Ancillary Services Unbundling
distribution network, in: Proceedings of the 22nd International Conference and Electricity Products - an Emerging Market, Tech. Rep. February (2004).
Exhibition on Electricity Distribution (CIRED 2013), no. 1241, Institution of [44] Braun M. Provision of Ancillary Services by Distributed Generators, [Ph.D. thesis],
Engineering and Technology, 2013, p. 4.〈http://dx.doi.org/10.1049/cp.2013. Kassel University (2008).
1137〉. [45] Nagel. H, Cichowski Hrsg. Rolf R., Systematische Netzplanung, 2nd Edition,
[15] Wieben E, Kumm T, Hohn E, Rohr M, Stadler M. The 5% approach as building Berlin : VDE-Verl.; Frankfurt, M. : VWEW-Energieverl., (2008).
block of an energy system dominated by renewables, in: Proceedings of the 28th [46] Pilo F, Jupe S, Silvestro F, Bakari KE, Abbey C. Planning and Optimization
Conference on Environmental Informatics - Informatics for Environmental Methods for Active Distribution Systems, Tech. Rep. August, CIGRE (2014).
Protection, Sustainable Development and Risk Management, BIS-Verlag, [47] ETG-Task Force Aktive EnergieNetze, Aktive Energienetze im Kontext der
Oldenburg, 2014, pp. 85–92. Energiewende: Anforderungen an künftige Übertragungs- und Verteilungsnetze
[16] Federal Network Agency (BNetzA), Monitoringbericht 2014, Tech. rep., unter Berücksichtigung von Marktmechanismen, Tech. rep., Energietechnische
Bundesnetzagentur, Berlin (2014). Gesellschaft im VDE (ETG) (ed.) (2013).
[17] German Energy Agency (dena), dena-Verteilnetzstudie. Ausbau- und [48] J. Schlabbach, K.-H. Rofalski, Power System Engineering, Wiley-VCH Verlag
Innovationsbedarf der Stromverteilnetze in Deutschland bis 2030. (2012). GmbH & Co. KGaA, Weinheim, Germany, (2008), 〈http://dx.doi.org/10.1002/
[18] Wlodarczyk P, Sumper A, Cruz M. Voltage control of distribution grids with multi- 9783527622795〉.
microgrids using reactive. Power Manag, Adv Electr Comput Eng [49] W. Kaufmann, Planung öffentlicher Elektrizitätsverteilungs-Systeme, VDE-Verlag,
2015;15(1):83–8. http://dx.doi.org/10.4316/AECE.2015.01012. 1995.
[19] Leemput N, Geth F, van Roy J, Olivella-Rosell P, Driesen J, Sumper A. MV and LV [50] Willis HL. Power Distribution Planning Reference Book, 2nd Edition, CRC Press, 2004.
residential grid impact of combined slow and fast charging of electric vehicles. [51] Deutsche Gesellschaft für Sonnenenergie e.V., EEG Anlagenregister, 〈http://www.
Energies 2015;8(3):1760–83. http://dx.doi.org/10.3390/en8031760. energymap.info〉, [Accessed 20 May 15].
[20] Resch M, Ramadhani B, Bühler J, Sumper A. Comparison of control strategies of [52] Nykamp S. Integrating Renewables in Distribution Grids : Storage, regulation and
residential PV storage systems, in: Proceedings of the 9th International Renewable the interaction of different stakeholders in future grids (2013). 〈http://dx.doi.org/
Energy Storage Conference (IRES 2015), 2015, p. 18. 〈http://dx.doi.org/10. 10.3990/1.9789036500579〉.
13140/RG.2.1.3668.2084〉. [53] Wirth G. Modellierung der Netzeinflüsse von Photovoltaikanlagen unter
[21] Camacho-Rascon C, Resch M, Schachler B, Bühler J, Sumper A. Increasing the Verwendung meteorologischer Parameter. [Ph.D. thesis], Carl von Ossietzky
hosting capacity of distribution grids by implementing residential PV storage Universität Oldenburg (2014).
systems and reactive power control, in: Proceedings of the 2016 13th International [54] Pardatscher R, Witzmann R, Wirth G, Becker G, Garhamer M, Brandtl J. Research
Conference on the European Energy Market (EEM), IEEE, 2016, pp. 1–5. 〈http:// on the impact of photovoltaic power generation in low and medium voltage grids.
dx.doi.org/10.1109/EEM.2016.7521338〉. in: Internationaler ETG-Kongress 2011, Würzburg, (2011).
[22] Díaz-González F, Sumper A, Gomis-Bellmunt O, Villafáfila-Robles R. A review of [55] Resch M, Bühler J, Huyskens H, Sumper A. Optimale Positionierung von
energy storage technologies for wind power applications. Renew Sustain Energy Großbatterien in Verteilnetzen. in: 30. Symposium Photovoltaische Solarenergie,
Rev 2012;16(4):2154–71. http://dx.doi.org/10.1016/j.rser.2012.01.029. OTTI e.V., 2015, p. 37, 〈http://dx.doi.org/10.13140/RG.2.1.1308.9123〉.
[23] Federal Ministry of Justice and Consumer Protection, Law on electricity and gas [56] Büchner J, Flörcken O, Dierkes S, Verheggen L, Uslar M. Moderne Verteilernetze
supply (Energiewirtschaftsgesetz-EnWG) (2015). für Deutschland. Tech. Rep. 44, BMWi (2014).
[24] Association for Electrical; Electronic & Information Technology (VDE), Power [57] Agora Energiewende, Stromspeicher in der Energiewende. Tech. rep., Agora
transformers - Part 2: Temperature rise for liquid-immersed transformers (IEC Energiewende, Berlin (2014).
60076-2:2011); German version EN 60076-2:2011, Tech. rep. (2012). [58] Liebenau V, Schwippe J, Kuch S, Rehtanz C. Network extension planning
[25] Association for Electrical; Electronic & Information Technology (VDE), Power considering the uncertainty of feed-in from renewable energies. 2013 IEEE
cables - Part 603: Distribution cables of rated voltage 0,6/1 kV; German version Grenoble Conference (2013) 1–6, 〈http://dx.doi.org/10.1109/PTC.2013.6652382〉.
HD 603 S1:1994/A3:2007, parts 0, 1, 3-G and 5-G (2010). [59] Schmidtner T. Probabilistische Methoden in der Netzplanung Niederspannung, in:
[26] German Institute for Standardisation (DIN), Voltage characteristics of electricity VDE-Kongress 2012 - Intelligente Energieversorgung der Zukunft, VDE-Verlag,
supplied by public distribution networks; German version EN 50160: 2010+ Cor.: 2012, p. 6.
2010 (2011). [60] Neimane V. On development planning of electricity distribution networks. [Ph.D.
[27] Association of German Grid Operators (VDN), TransmissionCode 2007 - Netz- thesis], KTH (2001).
und Systemregeln der deutschen Übertragungsnetzbetreiber, Tech. rep. (2007). [61] Jan K. Agent-based Simulation Environment for Improving the Planning of
[28] Kerber G, Witzmann R. Loading Capacity of Standard Oil Transformers on Distribution Grids. Phd, Technischen Universität Dortmund (2014).
Photovoltaic Load Profiles, in: 10th World Renewable Energy Congress and [62] Stetz T, Diwold K, Kraiczy M, Geibel D, Schmidt S, Braun M. Techno-economic
Exhibition, 2008, pp. 1198–1203. assessment of voltage control strategies in low voltage grids. IEEE Trans Smart
[29] Labed M, Brand M, Rose H. A cost-effective approach for the grid integration of Grid 2014;5(4):2125–32. http://dx.doi.org/10.1109/TSG.2014.2320813.
distributed renewable resources. Int J Emerg Technol Adv Eng 2014;4(3):249–52. [63] Idlbi B, Diwold K, Stetz T, Wang H, Braun M. Cost-benefit analysis of central and
[30] Ackerman T, Koch M, Rothfuchs H, Martens N, Brown T. Verteilnetzstudie local voltage control provided by distributed generators in MV networks. in:
Rheinland-Pfalz, Tech. rep. (2014). Proceeding of the 2013 IEEE Grenoble Conference, (2013), pp. 1–6, 〈http://dx.
[31] Engel B, Laudahn S, Marggraf O, Schnettler A. Vergleich von technischer doi.org/10.1109/PTC.2013.6652333〉.
Wirksamkeit sowie Wirtschaftlichkeit zeitnah verfügbarer Verfahren zur [64] Idlbi B, Scheidler A, Stetz T, Braun M. Preemptive network reinforcement at LV
Sicherung der statischen Spannungshaltung in Niederspannungsnetzen mit level considering uncertainty in prediction of PV penetration scenarios. in: 2015
starker dezentraler Einspeisung, Tech. rep., TU Braunschweig, RWTH Aachen, TU IEEE Eindhoven PowerTech, IEEE, 2015, pp. 1–6, 〈http://dx.doi.org/10.1109/
München, FGH (2014). PTC.2015.7232793〉.
[32] German Association of the Energy and Water Industry (BDEW), Technische [65] Keane A, Ochoa LF, Borges CLT, Ault GW, Alarcon-Rodriguez AD, Currie RaF,
Richtlinie Erzeugungsanlagen am Mittelspannungsnetz (2008). Pilo F, Dent C, Harrison GP. State-of-the-Art techniques and challenges ahead for
[33] Association for Electrical; Electronic & Information Technology (VDE), VDE-AR- distributed generation planning and optimization. IEEE Trans Power Syst
N 4105 Generators connected to the low-voltage distribution network - Technical 2013;28(2):1493–502. http://dx.doi.org/10.1109/TPWRS.2012.2214406.
requirements for the connection to and parallel operation with low-voltage [66] Ganguly S, Sahoo NC, Das D. Recent advances on power distribution system
distribution networks (2011). planning A state-of-the-art survey. Energy Syst 2013;4(2):165–93. http://
[34] ENTSO-E, Requirements for Grid Connection Applicable to all Generators (2013). dx.doi.org/10.1007/s12667-012-0073-x.
[35] European Committee for Electrotechnical Standardization (CENELEC), [67] Georgilakis PS, Hatziargyriou ND. A review of power distribution planning in the
Requirements for generating plants to be connected in parallel with distribution modern power systems era Models, methods and future research. Electr Power
networks - Part 1: Connection to a LV distribution network above 16 A (2015). Syst Res 2015;121:89–100. http://dx.doi.org/10.1016/j.epsr.2014.12.010.
[36] European Committee for Electrotechnical Standardization (CENELEC), [68] Rezaee Jordehi A. Optimisation of electric distribution systems A review. Renew
Requirements for generating plants to be connected in parallel with distribution Sustain Energy Rev 2015;51:1088–100. http://dx.doi.org/10.1016/
networks - Part 2: Connection to a MV distribution network (2015). j.rser.2015.07.004.
[37] Forum Network Technology / Network Operation in the VDE (FNN), Anschluss [69] Alarcon-Rodriguez A, Ault G, Galloway S. Multi-objective planning of distributed
und Betrieb von Speichern am Niederspannungsnetz (2014). energy resources a review of the state-of-the-art. Renew Sustain Energy Rev
[38] Schwaegerl C, Bollen MHJ, Karoui K, Yagmur A. Voltage control in distribution 2010;14(5):1353–66. http://dx.doi.org/10.1016/j.rser.2010.01.006.
systems as a limitation of the hosting capacity for distributed energy resources, in: [70] Viral R, Khatod D. Optimal planning of distributed generation systems in
Proceedings of the 18th International Conference And Exhibition on Electricity distribution system a review. Renew Sustain Energy Rev 2012;16(7):5146–65.
Distribution (CIRED 2005): , IET, Turin, 2005, pp. 6–9. http://dx.doi.org/10.1016/j.rser.2012.05.020.
[39] Papathanassiou S, Hatziargyriou N, Anagnostopoulos P, Aleixo L. Capacity of [71] Koopmann S, Scheufen M, Schnettler A. Integration of stationary and transpor-
Distribution Feeders for Hosting DER (2014). table storage systems into multi-stage expansion planning of active distribution
[40] German Energy Agency (dena), dena-Studie Systemdienstleistungen 2030. grids. in: IEEE PES ISGT Europe 2013, IEEE, (2013), pp. 1–5, 〈http://dx.doi.org/
Sicherheit und Zuverlässigkeit einer Stromversorgung mit hohem Anteil er- 10.1109/ISGTEurope.2013.6695339〉.
neuerbarer Energien, Tech. rep. (2014). [72] Wong S, Bhattacharya K, Fuller J. Electric power distribution system design and
[41] Stetz T. Autonomous Voltage Control Strategies in Distribution Grids with planning in a deregulated environment. IET Gener Transm Distrib
Photovoltaic Systems: Technical and Economic Assessment, [Ph.D. thesis], 2009;3(12):1061. http://dx.doi.org/10.1049/iet-gtd.2008.0553.

1060
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

[73] El-Khattam W, Hegazy Y, Salama M. An integrated distributed generation [100] Poullikkas A. A comparative overview of large-scale battery systems for electricity
optimization model for distribution system planning. IEEE Trans Power Syst storage. Renew Sustain Energy Rev 2013;27:778–88. http://dx.doi.org/10.1016/
2005;20(2):1158–65. http://dx.doi.org/10.1109/TPWRS.2005.846114. j.rser.2013.07.017.
[74] Zou K, Agalgaonkar AP, Muttaqi KM, Perera S. Distribution system planning with [101] Soloveichik GL. Battery technologies for large-scale stationary energy storage.
incorporating DG reactive capability and system uncertainties. IEEE Trans Sustain Annu Rev Chem Biomol Eng 2011;2(1):503–27. http://dx.doi.org/10.1146/
Energy 2012;3(1):112–23. http://dx.doi.org/10.1109/TSTE.2011.2166281. annurev-chembioeng-061010-114116.
[75] Vahidinasab V. Optimal distributed energy resources planning in a competitive [102] Corey G, Eyer J. Energy Storage for the Electricity Grid : Benefits and Market
electricity market Multiobjective optimization and probabilistic design. Renew Potential Assessment Guide A Study for the DOE Energy Storage Systems
Energy 2014;66:354–63. http://dx.doi.org/10.1016/j.renene.2013.12.042. Program. Tech. Rep. February, Sandia National Laboratories, New Mexico (2010).
[76] Haffner S, Pereira L, Pereira L, Barreto L. Multistage model for distribution [103] Moseley PT, Garche J. Electrochemical Eenrgy Storage for Renewable Sources and
expansion planning with distributed generation - Part II numerical results. IEEE Grid Balancing, 2015. 〈http://dx.doi.org/10.1016/B978-0-444-62616-5.00010-
Trans Power Deliv 2008;23(2):924–9. http://dx.doi.org/10.1109/ 3〉.
TPWRD.2008.917911. [104] Battke B. Multi-purpose technologies, lock-in and efficiency - Policy implications
[77] Haffner S, Pereira LL, Pereira LL, Barreto L. Multistage Model for Distribution from the case of stationary electricity storage. [Ph.D. thesis], ETH Zürich (2014).
Expansion Planning With Distributed Generation -Part I Problem Formulation. [105] Beck H-P, Engel B, Hofmann L, Menges R, Turek T, Weyer H. Eignung von
IEEE Trans Power Deliv 2008;23(2):915–23. http://dx.doi.org/10.1109/ Speichertechnologien zum Erhalt der Systemsicherheit. Tech. rep., Energie-
TPWRD.2008.917916. Forschungszentrum Niedersachsen, Gosslar (2013).
[78] Naderi E, Seifi H, Sepasian MS. A Dynamic Approach for Distribution System [106] Fuchs G, Lunz B, Leuthold M, Sauer DU. Technology Overview on Electricity
Planning Considering Distributed Generation. IEEE Trans Power Deliv Storage. Tech. Rep. June (2012).
2012;27(3):1313–22. http://dx.doi.org/10.1109/TPWRD.2012.2194744. [107] Genoese F. Modellgestützte Bedarfs- und Wirtschaftlichkeitsanalyse von
[79] Tancredo Borges CL, Ferreira Martins V. Multistage expansion planning for active Energiespeichern zur Integration erneuerbarer Energien in Deutschland. [Ph.D.
distribution networks under demand and Distributed Generation uncertainties. thesis], Karlsruhe (2013).
Int J Electr Power Energy Syst 2012;36(1):107–16. http://dx.doi.org/10.1016/ [108] Ibrahim H, Beguenane R, Merabet A. Technical and financial benefits of electrical
j.ijepes.2011.10.031. energy storage. in: Electrical Power and Energy Conference, EPEC 2012, 2012, pp.
[80] Bagheri A, Monsef H, Lesani H. Integrated distribution network expansion 86–91. 〈http://dx.doi.org/10.1109/EPEC.2012.6474985〉.
planning incorporating distributed generation considering uncertainties, reliabil- [109] Consentec GmbH, Description of load-frequency control concept and market for
ity, and operational conditions. Int J Electr Power Energy Syst 2015;73:56–70. control reserves. Tech. rep. (2014).
http://dx.doi.org/10.1016/j.ijepes.2015.03.010. [110] Sioshansi R, Denholm P, Jenkin T. Market and policy barriers to deployment of
[81] Falaghi H, Singh C, Haghifam M-R, Ramezani M. DG integrated multistage energy storage. Econ Energy Environ Policy 2012;1(2):47–64. http://dx.doi.org/
distribution system expansion planning. Int J Electr Power Energy Syst 10.5547/2160-5890.1.2.4.
2011;33(8):1489–97. http://dx.doi.org/10.1016/j.ijepes.2011.06.031. [111] Sterner M, Thema M, Eckert F, Moser A, Schäfer A, Drees T. Stromspeicher in der
[82] Martins VF, Borges CLT. Active distribution network integrated planning incor- Energiewende. Tech. rep., Agora Energiewende, Berlin (2014).
porating distributed generation and load response uncertainties. IEEE Trans [112] Akhil, AA, Huff G, Currier AB, Kaun BC, Rastler DM, Chen SB, Bradshaw DT,
Power Syst 2011;26(4):2164–72. http://dx.doi.org/10.1109/ Gauntlett WD. Electricity storage handbook. Tech. Rep. July (2013). http://dx.
TPWRS.2011.2122347. doi.org/10.0/SAND2013-5131.
[83] Sedghi M, Aliakbar-Golkar M, Haghifam M-R. Distribution network expansion [113] Spiecker S, Vogel P, Weber C. Ökonomische Bewertung von Netzengpässen und
considering distributed generation and storage units using modified PSO algo- Netzinvestitionen, uwf 17 (2009) 321–331.
rithm. Int J Electr Power Energy Syst 2013;52(0):221–30. http://dx.doi.org/ [114] Frontier Economics, Consentec GmbH, Relevance of established national bidding
10.1016/j.ijepes.2013.03.041. areas for European power market integration - an approach to welfare oriented
[84] Chen H, Wang Z, Yan H, Zou H, Luo B. Integrated planning of distribution evaluation. Tech. Rep. October (2011).
systems with distributed generation and demand side response. Energy Procedia [115] Hirth L, Ziegenhagen I. Balancing power and variable renewables: Three links.
2015;75(51322702):981–6. http://dx.doi.org/10.1016/j.egypro.2015.07.314. Renew Sustain Energy Rev 2015;50:1035–51. http://dx.doi.org/10.1016/
[85] Saboori H, Hemmati R, Abbasi V. Multistage distribution network expansion j.rser.2015.04.180.
planning considering the emerging energy storage systems. Energy Convers [116] Narula CK, Martinez R, Onar O, Starke MR, Andrews G. Economic Analysis of
Manag 2015;105:938–45. http://dx.doi.org/10.1016/j.enconman.2015.08.055. Deploying Used Batteries in Power Systems. Tech. rep., Oak ridge national
[86] Kaempf E, Braun M. Expert Systems as Support to Strategic Network Planning. laboratory (2011).
Tech. rep., Fraunhofer IWES Kassel (2015), 〈http://dx.doi.org/10.13140/RG.2.1. [117] Zentralverband Elektrotechnik und Elektronikindustrie e. V. (ZVEI),
3473.7760〉. Energieeinsparung durch Blindleistungskompensation. Tech. rep., 〈http://www.
[87] Khator S, Leung L. Power distribution planning: a review of models and issues. in: zvei.org/Publikationen/Blindleistung.pdf〉, [accessed 30.09.14].
IEEE Transactions on Power Systems, Vol. 12, (1997), pp. 1151–1159. 〈http://dx. [118] Battke B, Schmidt TS, Grosspietsch D, Hoffmann VH. A review and probabilistic
doi.org/10.1109/59.630455〉. model of lifecycle costs of stationary batteries in multiple applications. Renew
[88] Gorien T. Distribution-system planning using mixed-integer programming 128 (2) Sustain Energy Rev 2013;25:240–50.
(1981) 70–79. [119] Aschenbrenner S. Untersuchung verschiedener Geschäftsmodelle für den Einsatz
[89] Abapour S, Zare K, Mohammadi-ivatloo B. Dynamic planning of distributed eines Batteriespeichers im Niederspannungsnetz - Theoretische Betrachtungen
generation units in active distribution network 9 (2015) 1455–1463, 〈http://dx. und Bestimmung von Kenngrößen, Masterthesis, Hochschule Regensburg (2013).
doi.org/10.1049/iet-gtd.2014.1143〉. [120] Malhotra A, Battke B, Beuse M, Stephan A, Schmidt T. Use cases for stationary
[90] Partanen J. A modified dynamic programming algorithm for sizing, locating and battery technologies: A review of the literature and existing projects. Renew Sustain
timing of feeder reinforcements. IEEE Trans Power Deliv 1990;5(1):277–83. Energy Rev 2016;56:705–21. http://dx.doi.org/10.1016/j.rser.2015.11.085.
http://dx.doi.org/10.1109/61.107285. [121] Deparment of Energy (DOE), Global Energy Storage Database, 〈http://www.
[91] Barin A, Pozzatti LF, Canha LN, Machado RQ, Abaide AR, Arend G. Multi- energystorageexchange.org〉, [Accessed 24 July 15].
objective analysis of impacts of distributed generation placement on the opera- [122] Klausen M, Resch M, Bühler J. Analysis of potential single and combined business
tional characteristics of networks for distribution system planning. Int J Electr models for stationary battery storage systems. Energy Procedia 2016;99:321–31.
Power Energy Syst 2010;32(10):1157–64. http://dx.doi.org/10.1016/ http://dx.doi.org/10.1016/j.egypro.2016.10.122.
j.ijepes.2010.06.015. [123] Toffler A. The Third Wave, 1980.
[92] Engels K, Haubrich H.-J. Probabilistic evaluation of voltage stability in MV [124] Grijalva S, Tariq MU. Prosumer-based smart grid architecture enables a flat,
networks. in: 2000 Power Engineering Society Summer Meeting (Cat. No. sustainable electricity industry. in: ISGT 2011, IEEE, 2011, pp. 1–6, 〈http://dx.
00CH37134), Vol. 4, IEEE, 2000, pp. 2075–2080. doi.org/10.1109/ISGT.2011.5759167〉.
[93] Koopmann S, Nicolai S, Schnettler A. Multifunctional operation of a virtual power [125] Moshövel J, Kairies K-P, Magnor D, Leuthold M, Bost M, Gährs S, Szczechowicz E,
plant in an active distribution grid: Modelling approaches and first field test Cramer M, Sauer DU. Analysis of the maximal possible grid relief from PV-peak-
experiences from the SmartRegion Pellworm project. in: IEEE PES Innovative power impacts by using storage systems for increased self-consumption. Appl
Smart Grid Technologies, Europe, IEEE, 2014, pp. 1–6, 〈http://dx.doi.org/10. Energy 2015;137:567–75. http://dx.doi.org/10.1016/j.apenergy.2014.07.021.
1109/ISGTEurope.2014.7028949〉. [126] Short W, Packey DJ. A Manual for the Economic Evaluation of Energy Efficiency
[94] Union of the Electricity Industry (EURELECTRIC), Decentralised Storage : Impact and Renewable Energy Technologies. Tech. Rep. March (1995).
on Future Distribution Grids. Tech. Rep. june (2012). [127] Ueckerdt F, Hirth L, Luderer G, Edenhofer O. System LCOE What are the costs of
[95] Sterner M, Eckert F, Thema M, Bauer F. Der positive Beitrag dezentraler variable renewables?. Energy 2013;63:61–75. http://dx.doi.org/10.1016/j.en-
Batteriespeicher für die stabile Stromversorgung. Tech. rep., FNES, OTH ergy.2013.10.072.
Regensburg (2015). [128] Ondraczek J, Komendantova N, Patt A. WACC the dog The effect of financing costs
[96] Forum Network Technology / Network Operation in the VDE (FNN), Connecting on the levelized cost of solar PV power. Renew Energy 2015;75:888–98. http://
and operating storage units in low voltage networks (2013). dx.doi.org/10.1016/j.renene.2014.10.053.
[97] TÜV Süd, Grid compabitility certified by TÜV SÜD (2015). [129] Branker K, Pathak M, Pearce J. A review of solar photovoltaic levelized cost of
[98] Kairies K-P, Magnor D, Sauer DU. Scientific measuring and evaluation program electricity. Renew Sustain Energy Rev 2011;15(9):4470–82. http://dx.doi.org/
for photovoltaic battery systems(WMEP PV-Speicher). Energy Procedia 10.1016/j.rser.2011.07.104.
2015;73:200–7. http://dx.doi.org/10.1016/j.egypro.2015.07.672. [130] Pawel I. The cost of storage - how to calculate the levelized cost of stored energy
[99] Ibrahim H, Ilinca A, Perron J. Energy storage systems - Characteristics and (LCOE) and applications to renewable energy generation. Energy Procedia
comparisons. Renew Sustain Energy Rev 2008;12(5):1221–50. http://dx.doi.org/ 2014;46:68–77. http://dx.doi.org/10.1016/j.egypro.2014.01.159.
10.1016/j.rser.2007.01.023. [131] Breyer C, Gerlach A. Global overview on grid-parity. Prog Photovolt: Res Appl

1061
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

2013;21(1):121–36. http://dx.doi.org/10.1002/pip.1254. 2014;5(2):974–81. http://dx.doi.org/10.1109/TSG.2013.2281175.

[132] Vartiainen Eero, Masson G, Breyer C. PV LCOE in Europe 2014-30. Tech. rep., [159] von Appen J, Stetz T, Braun M, Schmiegel A. Local voltage control strategies for
European PV Technology Platform Steering Commitee (2015), 〈http://dx.doi.org/ pv storage systems in distribution grids. IEEE Trans Smart Grid
10.13140/RG.2.1.4669.5520〉. 2014;5(2):1002–9. http://dx.doi.org/10.1109/TSG.2013.2291116.
[133] International Renewable Energy Agency (IRENA), Battery Storage for Renewables [160] Armstorfer A, Müller H, Biechl H, Alt B, Sollacher R, Most D, Szabo A, Köberle R,
: Market Status and Technology Outlook. Tech. Rep. January, IRENA (2015). Fiedeldey M. Operation of battery storage systems in smart grids. in:
[134] Horiba T. Lithium-Ion Battery Systems. Proc IEEE 2014;102(6):939–50. http:// Internationaler ETG-Kongress 2013, (2013), pp. 1–8.
dx.doi.org/10.1109/JPROC.2014.2319832. [161] Riesen Y, Ding P, Monnier S, Wyrsch N, Balli C. Peak Shaving Capability of
[135] Nykvist B, Nilsson M. Rapidly falling costs of battery packs for electric vehicles. Household Grid-Connected PV-System with Local Storage: A Case Study. In:
Nat Clim Change 2015;5(4):329–32. http://dx.doi.org/10.1038/nclimate2564. Proceedings of the 28th European Photovoltaic Solar Energy Conference and
[136] Moreno B, López AJ, García-Álvarez MT. The electricity prices in the European Exhibition, (2013), pp. 3740–3744, 〈http://dx.doi.org/10.4229/
Union. The role of renewable energies and regulatory electric market reforms. 28thEUPVSEC2013-5CO.7.4〉.
Energy 2012;48(1):307–13. http://dx.doi.org/10.1016/j.energy.2012.06.059. [162] Siller M. Stromspeicher zur Netzstabilisierung. in: Bayerischer Energiekongress,
[137] Weron R. Electricity price forecasting A review of the state-of-the-art with a look 2013, p. 25.
into the future. Int J Forecast 2014;30(4):1030–81. http://dx.doi.org/10.1016/ [163] Schnettler A, Nilges J, Stolte A, Nykamp S, Smolka T, Matrose C, Willing S.
j.ijforecast.2014.08.008. Improving quality of supply and usage of assets in distribution grids by
[138] Ewi, Gws, Prognos, Entwicklung der Energiemärkte - Energiereferenzprognose. introducing a smart operator. In: Proceedings of the 22nd International
Tech. rep., Ewi, Gws, Prognos, Basel/Köln/Osnabrück (2014). Conference and Exhibition on Electricity Distribution (CIRED 2013), no. 0718,
[139] Talavera DL, de la Casa J, Muñoz-Cerón E, Almonacid G. Grid parity and self- 2013, pp. 0718–0718. 〈http://dx.doi.org/10.1049/cp.2013.0892〉.
consumption with photovoltaic systems under the present regulatory framework [164] Nykamp S, Molderink A, Hurink JL, Smit GJM. Storage operation for peak
in Spain: the case of the University of Jaén Campus. Renew Sustain Energy Rev shaving of distributed PV and wind generation. in: 2013 IEEE PES Innovative
2014;33:752–71. http://dx.doi.org/10.1016/j.rser.2014.02.023. Smart Grid Technologies Conference (ISGT), IEEE, (2013), pp. 1–6, 〈http://dx.
[140] Chiaroni D, Chiesa V, Colasanti L, Cucchiella F, Frattini F. Evaluating solar energy doi.org/10.1109/ISGT.2013.6497786〉.
profitability: A focus on the role of self-consumption. Energy Convers Manag [165] Goergens P, Potratz F, Matrose C, Schumann S, Schnettler A, Willing S, Smolka T.
2014;88:317–31. http://dx.doi.org/10.1016/j.enconman.2014.08.044. An online learning algorithm approach for low voltage grid management. In:
[141] Luthander R, Widén J, Nilsson D, Palm J. Photovoltaic self-consumption in Proceedings of the 22nd International Conference and Exhibition on Electricity
buildings: A review. Appl Energy 2015;142:80–94. http://dx.doi.org/10.1016/ Distribution (CIRED 2013), Vol. 5, (2013), pp. 0702–0702, 〈http://dx.doi.org/10.
j.apenergy.2014.12.028. 1049/cp.2013.0885〉.
[142] Hoppmann J, Volland J, Schmidt TS, Hoffmann VH. The economic viability of [166] Li J, Danzer MA. Optimal charge control strategies for stationary photovoltaic
battery storage for residential solar photovoltaic systems - A review and a battery systems. J Power Sources 2014;258:365–73. http://dx.doi.org/10.1016/
simulation model. Renew Sustain Energy Rev 2014;39:1101–18. http:// j.jpowsour.2014.02.066.
dx.doi.org/10.1016/j.rser.2014.07.068. [167] Williams C, Binder J, Danzer M, Sehnke F, Felder M. Battery Charge Control
[143] Kost C, Mayer JN, Thomsen J, Hartmann N, Senkpiel C, Philipps S, Nold S, Lude Schemes for Increased Grid Compatibility of Decentralized PV Systems. In:
S, Schlegl T. Studie Stromgestehungskosten Erneuerbare Energien. Tech. Rep. Proceedings of the 28th European Photovoltaic Solar Energy Conference and
November, Fraunhofer-Institut für solare Energiesysteme ISE (2013). Exhibition, (2013), pp. 1–6.
[144] German Association of the Energy and Water Industry (BDEW), [168] Lödl M, Witzmann R, Metzger M. Operation strategies of energy storages with
Industriestrompreise. Tech. Rep. April, Bundesverband der Energie- und forecast methods in low-voltage grids with a high degree of decentralized
Wasserwirtschaft e.V. (BDEW), Berlin (2014). generation. 2011 IEEE Electrical Power and Energy Conference (2011) 52–56,
[145] Kairies K-P, Haberschusz D, Magnor D, Leuthold M, Sauer DU. 〈http://dx.doi.org/10.1109/EPEC.2011.6070252〉.
Wissenschaftliches Mess- und Evaluierungsprogramm Solarstromspeicher - [169] Zeh A, Witzmann R. Operational strategies for battery storage systems in low-
Jahresbericht 2015. Tech. rep., Institut für Stromrichtertechnik und Elektrische voltage distribution grids to limit the feed-in power of roof-mounted solar power
Antriebe RWTH Aachen (2015). systems. Energy Procedia 2014;46:114–23. http://dx.doi.org/10.1016/j.egy-
[146] Kästel P, Gilroy-Scott B. Economics of pooling small local electricity prosumers pro.2014.01.164.
-LCOE & self-consumption. Renew Sustain Energy Rev 2015;51:718–29. http:// [170] Niimura T, Ozawa K, Yamashita D, Yoshimi K, Osawa M. Profiling residential PV
dx.doi.org/10.1016/j.rser.2015.06.057. output based on weekly weather forecast for home energy management system. in:
[147] Arghandeh R, Woyak J, Onen A, Jung J, Broadwater RP. Economic optimal 2012 IEEE Power and Energy Society General Meeting, IEEE, 2012, pp. 1–5,
operation of community energy storage systems in competitive energy 〈http://dx.doi.org/10.1109/PESGM.2012.6345020〉.
markets. Appl Energy 2014;135:1–17. http://dx.doi.org/10.1016/j.ape- [171] Lorenz E, Hurka J, Heinemann D, Beyer HG. Irradiance Forecasting for the Power
nergy.2014.08.066. Prediction of Grid-Connected Photovoltaic Systems. IEEE J Sel Top Appl Earth
[148] Nourai A, Sastry R, Walker T. A vision & strategy for deployment of energy Obs Remote Sens 2009;2(1):2–10. http://dx.doi.org/10.1109/
storage in electric utilities. in: IEEE PES General Meeting, IEEE, 2010, pp. 1–4, JSTARS.2009.2020300.
〈http://dx.doi.org/10.1109/PES.2010.5589368〉. [172] Chicco G, Cocina V, Di Leo P, Spertino F. Weather forecast-based power predictions
[149] Gaudchau E, Resch M, Zeh A. Quartierspeicher: Definition, rechtlicher Rahmen and experimental results from photovoltaic systems. in: 2014 International
und Perspektiven. Ökologisches Wirtschaften 2016 (2) (2016) 26–27, 〈http://dx. Symposium on Power Electronics, Electrical Drives, Automation and Motion, IEEE,
doi.org/10.14512/OEW310226〉. 2014, pp. 342–346, 〈http://dx.doi.org/10.1109/SPEEDAM.2014.6872086〉.
[150] Parra D, Gillott M, Norman SA, Walker GS. Optimum community energy storage [173] Bergner J, Weniger J, Tjaden T, Quaschning V. Feed-in Power Limitation of Grid-
system for PV energy time-shift. Appl Energy 2015;137:576–87. http:// Connected PV Battery Systems with Autonomous Forecast-Based Operation
dx.doi.org/10.1016/j.apenergy.2014.08.060. Strategies. In: Proceedings of the 29th European PV Solar Energy Conference and
[151] Zeh A, Rau M, Witzmann R. Comparison of decentralised and centralised grid- Exhibition, Amsterdam, 2014.
compatible battery storage systems in distribution grids with high PV penetration. [174] Evans KF. The spherical harmonics discrete ordinate method for three-dimen-
in: Progress in Photovoltaics: Research and Applications, (2014), p. 11, 〈http:// sional atmospheric radiative transfer. J Atmos Sci 1998;55(3):429–46.
dx.doi.org/10.1002/pip.2566〉. [175] Dierer S, Remund J, Cattin R, Koller T, Strasser P. BFE, Einspeiseprognosen für
[152] Weniger J, Tjaden T, Quaschning V. Sizing and grid integration of residential PV neue erneuerbare Energien. Tech. rep., Bundesamt für Energie BFE (2010).
battery systems. In: Proceedings of the 8th International Renewable Energy [176] The Weather Research & Forecasting, 〈http://www.wrf-model.org/〉, [Accessed
Storage Conference and Exhibition (IRES 2013), (2013). 01 October 2015].
[153] Moehrke F, Grueger F, Triebke H, Arnhold O, Myrzik J. Model-based quantifi- [177] Meteotest, 〈http://www.meteotest.ch/〉, [Accessed 01 October 2015].
cation of a microgrid via key performance indicators. In: Proceedings of the 9th [178] Pelland S, Remund J, Kleissl J, Oozeki T, De Brabandere K. Photovoltaic and solar
International Renewable Energy Storage Conference (IRES 2015), no. March forecasting: state of the art. Tech. rep., International energy agency (IEA) (2013).
(2015), p. 11. [179] Luchscheider P. Smart Grid Solar: A Bavarian Smart Energy Project (2015).
[154] Thomann R, Schubert M. Strombank - Innovatives Betreibermodell für [180] Fünfgeld C, Tiedemann R. Anwendung der Repräsentativen VDEW-Lastprofile
Quartierspeicher. in: Staatskolloquium Umweltforschung Baden-Württemberg, step by step. VDEW Materialien M-05/2000 (2000) 34.
(2015), p. 12. [181] European Network of Transmission System Operators (ENTSO-E, Appendix 1 -
[155] Struth J, Kairies K-P, Leuthold M, Aretz A, Hirschl PB, Schnettler PA, Dirk P, Load-Frequency Control and Performance, Tech. rep. (2009).
Sauer DU, Bost M, Gährs S, Cramer M, Szczechowicz E. PV-Benefit: a Critical [182] UCTE, Operation Handbook. Tech. rep., UCTE (2004).
Review of the Effect of Grid Integrated PV-Storage-Systems. in: Iternational [183] Díaz-González F, Hau M, Sumper A, Gomis-Bellmunt O. Participation of wind
Renewable Energy Storage Conference (IRES 2013), (2013), p. 10. power plants in system frequency control: review of grid code requirements and
[156] Ueda Y, Kurokawa K, Kitamura K, Akanuma K, Yokota M, Sugihara H. Study on control methods. Renew Sustain Energy Rev 2014;34:551–64. http://dx.doi.org/
the overvoltage problem and battery operation for grid-connected residential PV 10.1016/j.rser.2014.03.040.
systems. In: Proceedings of the 22nd European Photovoltaic Solar Energy [184] Koller M, Borsche T, Ulbig A, Andersson G. Review of grid applications with the
Conference (September) (2007) 3094–3097. Zurich 1 MW battery energy storage system. Electr Power Syst Res
[157] Schneider M, Boras P, Schaede H, Quurck L, Rinderknecht S. Effects of 2015;120:128–35. http://dx.doi.org/10.1016/j.epsr.2014.06.023.
Operational strategies on performance and costs of electric energy storage [185] Pape C, Gerhardt N, Härtel P, Scholz A, Schwinn R, Drees T, Maaz A, Sprey J,
systems. Energy Procedia 2014;46(Ires 2013):271–80. http://dx.doi.org/ Breuer C, Moser A, Sailer F, Reuter S, Müller T. Roadmap Speicher:
10.1016/j.egypro.2014.01.182. Speicherbedarf für Erneuerbare Energien - Speicheralternativen - Speicheranreiz -
[158] Marra F, Yang G, Traeholt C, Ostergaard J, Larsen E. A decentralized storage Überwindung rechtlicher Hemmnisse (Kurzzusammenfassung), Tech. rep., IWES,
strategy for residential feeders with photovoltaics. IEEE Trans Smart Grid IAEW, Stiftung Umweltenergierecht (2014).

1062
M. Resch et al. Renewable and Sustainable Energy Reviews 74 (2017) 1042–1063

[186] Thien T, Axelsen H, Merten M, Axelsen H, Zurmùhlen S, Leuthold M. Planning of files/NNVWV_Muster_1312_mit_Anlagen.pdf〉, [Accessed 30 September 2014].
Grid-Scale Battery Energy Storage Systems: Lessons Learned from a 5 MW [206] 50 Hertz, Preisblatt 2014, 〈http://www.50hertz.com/de/Anschluss-Zugang/
Hybrid Battery Storage Project in Germany. in: Battcon - International Stationary Netzzugang/Netznutzungspreis〉, [Accessed 30 September 2014].
Battery Conference 2015, (2015), p. 10. [207] Tennet, Netznutzungsvertrag 2012, 〈http://www.tennet.eu/de/fileadmin/
[187] Aundrup T, Beck H-P, Becker A, Berthold A. Batteriespeicher in der Nieder- und downloads/Kunden/nn_vertrag.PDF〉, [Accessed 30 September 2014].
Mittelspannungsebene - Anwendungen und Wirtschaftlichkeit sowie [208] Rubel H, Pieper C. Revisiting Energy Storage, Tech. rep., Boston Consulting
Auswirkungen auf die elektrischen Netze, Tech. rep., Energietechnischen Group, Frankfurt (2011).
Gesellschaft im VDE (ETG) (2015). [209] Böttger D, Bruckner T. Marktszenarien für eine erfolgreiche und nachhaltige
[188] Bühler J, Resch M, Wiemann J, Twele J. Lebenszyklusanalyse von Großbatterien Energiewende 11. in: Energieversorgungssysteme der Zukunft, Leipzig, (2014), p. 38.
am deutschen Regelenergiemarkt. in: 9. Internationale Energiewirtschaftstagung [210] USV System, Different UPS-systems, 〈http://www.online-usv.de/de/produkte/
(IEWT), (2015), p. 44. 〈http://dx.doi.org/10.13140/RG.2.1.4454.6400〉. intro.php〉, [accessed 20 May 2015].
[189] Gitis A, Thien T, Leuthold M, Sauer DU. Optimization of Battery Energy Storage [211] E.ON, E.ON Grundversorgung Strom (Doppeltarif), 〈http://www.eon.de/pk/de/
Systems for Primary Control Reserve. In: Proceedings of the 8th International strom/flexibelstrom/grundversorgung-strom/grundversorgung-strom-
Renewable Energy Storage Conference and Exhibition (IRES 2013), (2013), p. 8. doppeltarif.html〉, [Accessed 30 September 2014].
[190] Sasaki T, Kadoya T, Enomoto K. Study on load frequency control using redox flow [212] EnBW, Intelligente Stromzähler, 〈https://www.enbw.com/privatkunden/tarife-
batteries. IEEE Trans Power Syst 2004;19(1):660–7. http://dx.doi.org/10.1109/ und-produkte/strom/index.html/〉, [Accessed 29 September 2014].
TPWRS.2003.818724. [213] RWE, RWE EDL21 Meter Strom, 〈https://www.rwe.de/web/cms/de/1160044/
[191] Sauer DU, Lunz B, Magnor D. Marktanreizprogramm für dezentrale Speicher privatkunden/strom-vergleich/stromrechner/〉, [Accessed 29 September 2014].
insbesondere für PV-Strom, Tech. rep. (2013). [214] Vattenfall, Easy Spar Aktiv, 〈https://www.vattenfall.de/de/berlin-easy-spar-aktiv.
[192] Kondziella H, Brod K, Bruckner T, Olbert S, Mes F. Stromspeicher für die htm〉, [Accessed 30 September 2014].
Energiewende - eine akteursbasierte Analyse der zusätzlichen Speicherkosten, [215] Amprion, Entgelte, 〈http://www.amprion.net/sites/default/files/
Zeitschrift für Energiewirtschaft 37 (2013) 249–260 〈http://dx.doi.org/10.1007/ EntgelteAmprion2014demitUmlagen.pdf〉, [Accessed 30 September 2014].
s12398-013-0115-7〉. [216] Tennet, Netzentgelte, 〈http://www.tennet.eu/de/fileadmin/downloads/Kunden/
[193] Gawlik W. Speicher für die Energiewende, e & i Elektrotechnik und 131216_TenneT_Netzentgelte_fuer_2014_Deutsch.pdf〉, [Accessed 25
Informationstechnik 130 (8) (2013) 250–250 〈http://dx.doi.org/10.1007/ September 2014].
s00502-013-0163-z〉. [217] TransnetBW, Entgelte Netznutzung, 〈http://www.transnetbw.de/downloads/
[194] Auer J, Keil J. Moderne Stromspeicher - Unverzichtbare Bausteine der netzzugang/entgelt/Vorlaeufige-Preise-Netznutzung-2014.pdf〉, [Accessed 21
Energiewende, Tech. rep., Deutsche Bank (2012). March 2016].
[195] Moser A, Zdrallek M, Krause H, Graf F. Nutzen von Smart-Grid-Konzepten unter [218] German Association of the Energy and Water Industry (BDEW),
Berücksichtigung der Power-to-Gas-Technologie, Tech. rep., DVGW Deutscher Strompreisanalyse, Tech. rep., Berlin (2014).
Verein des Gas- und Wasserfaches e.V. (2014). [219] Roon S, Sutter M, Samweber F, Wachunger K. Netzausbau in Deutschland, Tech.
[196] German TSO, Anforderungen an die Speicherkapazität bei Batterien für die rep., Konrad-Adenauer-Striftung e.V. (2014).
Primärregelleistung (2015). [220] Leprich U. Netze, Speicher, Lastmanagement. in: BUND-Tagung: Welches
[197] German TSO, Eckpunkte und Freiheitsgrade bei Erbringung von Stromnetz braucht die Energiewende?, Stuttgart, 2014, p. 34.
Primärregelleistung - Leitfaden für Anbieter von Primärregelleistung (2014). [221] Thomann R, Markus S, Pickford G, Bossler A. Strombank - Innovatives
[198] Schürer T. Elektrische Speicher aus der Sicht eines Verteilnetzbetreibers. in: Betreibermodell für Quartierspeicher (Zwischenbericht), Tech. rep. (2014).
Fachforum: Energiespeicher im Kontext der Energiewende, IHK Akademie [222] Thomann R. Innovatives Betreibermodell für Quartierspeicher. in: Kongress
Mittelfranken, Nürnberg, (2015), p. 20. Energie- und Energiespeichertechnologien, no. November, Stuttgart, (2014), p.
[199] Hollinger R, Diazgranados LM, Erge T. Trends in the German PCR market: 17.
Perspectives for battery systems. In: 2015 Proceedings of the 12th International [223] 〈http://www.lew-verteilnetz.de/〉, [Accessed 01 October 2015].
Conference on the European Energy Market (EEM), IEEE, (2015), pp. 1–5, [224] Jossen A, Gasteiger H, Müller M. EEBatt - Dezentrale Stationäre Batteriespeicher
〈http://dx.doi.org/10.1109/EEM.2015.7216661〉. zur effizienten Nutzung Erneuerbarer Energien und Unterstützung der
[200] D'Adamo C, Buchholz B, Abbey C, Khattabi M, Jupe S, Pilo F. Development and Netzstabilität - Zwischenbericht 2015, Tech. rep. (2015).
operation of active distribution networks: Results of CIGRÉ C6.11 Working [225] Meyer I. Quartiersspeicher von IBC SOLAR als Baustein für die Energiewende in
group. In: CIRED - Proceedings of the 21st International Conference on Electricity Bayern - press release, 〈http://www.ibc-solar.de/uploads/media/150505-Epplas-
Distribution., no. June (2011). IBC-ZAE.pdf〉, [Accessed 15 May 2015].
[201] Weniger J, Bergner J, Beier D, Jakobi M, Tjaden T, Quaschning V. Grid Feed-in [226] Association of German Grid Operators (VDN), Transmission Code 2003 -
Behavior of Distributed Pv Battery Systems. In: Proceedings of the 31st European Unterlagen zur Präqualifikation für die Erbringung von Primärregelleistung für
PV Solar Energy Conference and Exhibition, Hamburg, (2015), pp. 1603–1606. die ÜNB, Tech. rep. (2003).
[202] Gönen T. Electric Power Distribution Engineering, 3rd edition. New York: CRC [227] Otte M, Patt J, Lück J. Beschluss BK6-10-097, Tech. rep., Bundesnetzagentur
Press; 2014. (2011).
[203] EPEX-SPOT, EPEX-SPOT-Data for 2013 (2014). [228] ZEIT ONLINE, Größter kommerzieller Batteriespeicher Europas in Schwerin am
[204] Regeleistungs.net, Ausschreibungsübersicht 2013, 〈https://www.regelleistung. Netz, 〈http://www.zeit.de/news/2014-09/16/deutschland-groesster-
net/ip/action/ausschreibung/public〉, [Accessed 30 September 2014]. kommerzieller-batteriespeicher-europas-in-schwerin-am-netz-16132813〉,
[205] Amprion, Netznutzungsvertrag 2013, 〈http://www.amprion.net/sites/default/ [accessed 16 September 2014].

1063