Agent-based Simulation of the German and French Wholesale
Electricity Markets
Recent Extensions of the PowerACE Model with Exemplary Applications
Andreas Bublitz, Philipp Ringler, Massimo Genoese and Wolf Fichtner
Chair of Energy Economics, Institute for Industrial Production (IIP), Karlsruhe Institute of Technology (KIT),
Hertzstraße 16, 76137 Karlsruhe, Germany
Keywords: Economic Agent Models, Auctions and Markets, Agent-based Simulation, Multi-agent Systems, Electricity
Markets.
Abstract: Given electricity markets’ complexity, model-based analysis has proven to be a valuable tool for decision
makers in related industries or politics. Among the different modelling techniques for electricity markets,
agent-based modelling offers specific advantages. In this paper, the detailed agent-based simulation model
for the wholesale electricity market, PowerACE, is presented with its latest extensions. The model integrates
the short-term perspective of daily electricity trading and long-term capacity expansion planning. Various
market elements are simulated including the day-ahead market as well as the coupling of different market
areas with limited interconnection capacities. Strategic behaviour of the main supply-side agents is taken
into account. The model has already been applied to various research questions regarding the development
of electricity markets and the behaviour of market participants. In this contribution, exemplary results for
the market coupling of the German and French wholesale electricity market are shown. In the future, due to
the current developments in the electricity markets, the PowerACE modelling framework is to be extended
by various aspects including the simulation of an intraday market and the integration of different aspects of
uncertainty which becomes necessary given current developments in the electricity markets.
1 INTRODUCTION
Today’s liberalized wholesale electricity markets are
generally considered to be highly complex systems.
This is due to, among other things, the specific
characteristics of the commodity electricity (e.g.
instantaneous balancing of supply and demand,
limited storability) and the fact that electricity can
only be transported by a transmission grid with
limited capacities. Other factors that increase the
complexity are the various interrelated markets
where electricity or related products can be traded
(e.g. day-ahead market, future market) and the
influence of other volatile markets such as the
market for carbon emission allowances.
One important aspect of electricity systems is the
reliability which should be ensured at all times. In
liberalized European markets electricity generation
companies are not obliged to invest in new power
plants. Consequently, electricity markets need to be
designed in such a way that there are sufficient
incentives for adequate investments. The currently
often discussed concept to ensure reliability in
Europe is called “energy only” because power plant
operators generate their profits mainly from the
produced energy but are not compensated for only
providing generation capacity that ensures
reliability.
In Germany and several other European
countries the spot market for electricity, in particular
the day-ahead market auctions organized by
electricity exchanges, plays an important role as it
provides a market place to sell or buy electricity and
its price serves as a reference for other markets (e.g.
future markets, bilateral contracts).
In addition,
reserve markets are implemented to ensure the short-
term reliability of the electricity system.
Two important developments currently altering
the economics of European electricity markets are
the increasing electricity generation from renewable
energy sources and the European market integration.
While for a long time mainly nuclear, coal and oil
power plants had been installed in Europe,
governments have recognized the decarbonisation
40
Bublitz A., Ringler P., Genoese M. and Fichtner W..
Agent-based Simulation of the German and French Wholesale Electricity Markets - Recent Extensions of the PowerACE Model with Exemplary
Applications.
DOI: 10.5220/0004760000400049
In Proceedings of the 6th International Conference on Agents and Artificial Intelligence (ICAART-2014), pages 40-49
ISBN: 978-989-758-016-1
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
potential of the electricity sector and there has been
a continuous trend to move towards renewables and
gas. Specifically, the introduction of the European
Union Emissions Trading System and the creation of
various policy programs to support the use of
renewable energy sources have contributed to this
development.
However, the feed-in of electricity generated
from photovoltaic and wind power poses challenges
to the electricity markets in their current form
because in comparison with thermal power plants
the generation from these sources is neither
projectable nor exactly predictable and typically
enjoys a guaranteed feed-in and compensation,
respectively. Consequently, operators of
conventional power plants are faced with another
source of uncertainty that needs to be considered
within the unit commitment problem, where an
optimal balance of demand and supply under the
various technical constraints of the power plants is
to be determined. After determining the day-ahead
operation schedule, the intraday market, where
electricity can be traded at short notice, offers a
possibility to adjust the schedule based on updated
information, e.g. forecast of renewable generation.
The intraday market is likely to gain importance in
the next years, as the generation from renewable
energy sources is expected to further increase.
Another important development in the electricity
market is that the current borders of the national
markets are subject to change; there are ongoing
efforts to achieve a single European market. One
aspect thereof is the implementation of market-based
mechanisms to allocate limited cross-border
capacities between European countries. The Central
Western Europe (CWE) Market Coupling between
Germany, France, Belgium, the Netherlands and
Luxemburg serves as one of the most prominent
examples. Market coupling maximizes social
welfare, leads to price convergence and helps to
balance different supply and demand situation in the
interconnected market areas. The integration of
markets is a matter-of-fact, thus influencing market
prices and profitability of power plants in Europe.
Given the electricity system’s complexity the
relevant actors rely on different types of models for
decision support. For instance, models are used by
regulatory entities to analyse questions related to
market design which is necessary to guarantee
system reliability on different levels. Similarly,
generation companies rely on electricity market
models, for example, in order to examine investment
cases. Naturally, market changes need to be reflected
appropriately in modelling techniques.
In this paper, the main elements of the detailed
bottom-up agent-based simulation model PowerACE
are described and current extensions to adjust the
model to relevant electricity market developments
are presented. The aims of this paper are to present a
comprehensive overview of the PowerACE
modelling framework for electricity markets and
how it can be applied to different research questions.
The paper is organized as follows: section 2
provides a brief overview of the different types of
electricity market models and shows the general
suitability of agent-based simulation in the context
of electricity markets. In section 3, the model’s main
elements with a focus on agents and markets are
described. Exemplary results are presented in
section 4. Finally, section 5 concludes with a
summary and an outlook.
2 MODELS FOR ELECTRICITY
MARKETS
The models used for electricity markets can be
classified into several categories. Ventosa et al.
(2005) identify three major categories in electricity
market modelling: optimization models, equilibrium
models and simulation models. Distinguishing
features include the mathematical structure, market
representation, computational tractability and main
applications.
While in Europe the liberalization of electricity
markets started in 1996, electricity market models
developed beforehand had been mostly optimizing
models incorporating the perspective of a single
planner, i.e. the government. Through the
liberalization, the integration of a market perspective
in models has gained importance, which brought
forth the development of alternative models such as
agent-based models that are able to adequately
represent the current market situation where not one
central decision maker is found, but several market
players pursue their individual goals. In general,
agent-based models which have been developed in
quite different disciplines can provide a flexible
environment which allows considering inter alia
learning effects, imperfect competition including
strategic behaviour and asymmetric information
among market participants (Tesfatsion, 2006).
Nowadays, there exists a large number of agent-
based electricity market models. Depending on the
research focus, the models in the literature will differ
from each other with respect to various criteria.
Each agent-based model features a certain agent
Agent-basedSimulationoftheGermanandFrenchWholesaleElectricityMarkets-RecentExtensionsofthePowerACE
ModelwithExemplaryApplications
41
definition and architecture which can include
several dimensions. In the first place, it is essential
to define conceptually what the “agent” represents in
the model. In the field of Agent-based
Computational Economics (ACE) agents generally
are defined as having a set of data and pre-defined
behavioural rules within a computationally
constructed world (Tesfatsion, 2006). Agent
architecture includes the design of specific agent
decision models including adaptive learning
algorithms. Market modelling is another large
building block of agent-based models. Given the
complex nature of electricity wholesale markets and
the electricity supply chain, different types of
horizontally and vertically integrated markets exist.
In order to analyse the existing interrelations
between markets, one has to consider these markets
with respect to their specific clearing rules.
Depending on the spatial coverage of the model,
coupling of interconnected areas might be
considered as well. Similarly, agent-based models
differ with regard to the time resolution as well as
time scale of the simulation. The latter aspects
includes, for instance, whether short-term behaviour
(e.g. spot market bidding strategies) and long-term
aspects (e.g. investment decisions) are jointly
considered. Another important aspect of electricity
market models is the representation of the electricity
system’s technical constraints (e.g. techno-economic
aspects of generation units, grid constraints.
Three comprehensive review papers showing the
large body of agent-based models for electricity
markets and their distinctive features are provided
by Guerci et al. (2010), Weidlich and Veit (2008),
and Sensfuß et al. (2007). These literature reviews
contain a comparison of the different existing
models including the model presented in this paper.
Generally, having an integrated agent and market
perspective, as well as a high degree of flexibility,
agent-based simulation models can be used for
detailed analyses of electricity markets and
interactions therein. Potential applications include
market power analysis or market design studies
while considering the feed-in from renewable energy
sources and integrated markets with respect to
products, time and region.
3 POWERACE MODEL
3.1 Model Overview
The development of the PowerACE model started in
2004 and since then the model has been
continuously extended and applied to various
research questions.
The subject of modelling is the electricity
wholesale market which is simulated for each hour
of a year. Originally, the model was designed for the
German market area. However, Europe’s electricity
markets are all liberalized and set up according to
the same fundamental principles. That is why
PowerACE can be used to simulate other European
market areas as well. Market areas are interpreted as
one “object” in the programming environment
featuring different market elements, agents and input
data. In order to simulate different market areas, the
respective object is instantiated repeatedly.
One of the key features of the model is the
integration of both short-term market developments
and long-term capacity expansion planning.
Thereby, interactions and feedback loops between
short-term and long-term output decisions are
considered. Decisions regarding the expansion of
capacity, i.e. whether to install a new power plant
are influenced by current and future developments in
the daily electricity trading as the main source of
income and vice versa.
The key modules are markets, electricity supply,
electricity demand and regulatory aspects. The main
players participating in the wholesale electricity
market are modelled individually; small companies
are represented in an aggregated form. Different
types of market participants are modelled as
different types of agents. Each agent takes over
certain roles, makes decisions based on specified
functions and either takes part in or sets rules for a
respective market. A simplified overview of the
model structure with two market areas is given in
Figure 1.
In the following sections, the focus is set on the
supply side, i.e. on generation companies which
have to decide on the short-term operation of their
existing power plants and on the investment in new
ones.
3.2 Short-term Bidding on Electricity
Markets
The short-term operation of power plants is
determined by the SupplyBidder agent. The agent
evaluates the different markets where energy or
capacity of thermal power plants can be offered and
determines the operation schedule and dispatch of
the plants. Within PowerACE the day-ahead market
is the main spot market. In accordance with the
current situation in Central Western Europe, every
SupplyBidder daily submits for each available power
ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence
42
Figure 1: Simplified structure of the PowerACE model.
plant electricity supply bids to the day-ahead market.
Besides for thermal power plants, also supply bids
for generation from renewable energy sources, e.g.
wind or biomass, are regarded. Since pumped-
storage units can produce or consume electricity,
they submit either buy or sell bids. The same applies
to the electricity exchange with market areas which
are not explicitly modelled. After receiving the bids
the DayAheadMarketAuctioneer determines a
uniform price for each hour of the next day
considering all submitted supply and demand bids.
SupplyBidders are faced with an economic
optimization problem, where the offered volume and
price of their power plants needs to be determined
and which is solved in several steps. Firstly, the
available capacity P
i,d
of a power plant i on a day d
needs to be determined. Power plants may not be
available at all for a given day due to unexpected
issues, e.g. start-up failure, or expected reasons, e.g.
maintenance. Since power plants act on other
markets (e.g. reserve market) as well, the reserved
capacity P
r,i,d
for these markets is not available
anymore for the day-ahead market bidding and
needs to be subtracted from the net electrical
capacity P
net,i
:
P
i,d
=
P
net,i
P
r,i,d
if plant i is available
on day d
0 otherwise
(1)
Secondly, the bid price is calculated. It consists of
three elements: variable costs, start-up costs and a
potential mark-up. Variable costs c
var,i,d
represent the
direct costs of producing one unit of electricity and
are determined by the fuel price p
fuel,i,d
, the power
plant’s net electrical efficiency η
i
, the price of CO
2
emission allowances p
CO2,d
, the CO
2
emission factor
of the fuel EF
fuel
and the costs for operation and
maintenance c
O&M,i
:
c
var,i,d
=
p
fuel,i,d
η
i
+
p
CO2,d
·EF
fuel
η
i
+c
O&M,i
(2)
Changing the mode of operation of power plants,
i.e. starting up or shutting down, causes additional
costs. Firstly, material is stressed mainly by
temperature changes reducing life expectancy;
secondly, for start-ups fuel is needed in order to
reach the operating temperature of a power plant.
When determining the bid price the costs from start-
up and shutdown processes as an intertemporal
restriction can be considered by power plant
operators. In the PowerACE model, this means that
for base load running power plants also lower
market prices are accepted in order to avoid shutting
down the power plant. In turn, start-up costs are
added to the bid price for peak load power plants in
order to earn start-up costs in hours where the plant
is expected to be running. To estimate start-up costs
Agent-basedSimulationoftheGermanandFrenchWholesaleElectricityMarkets-RecentExtensionsofthePowerACE
ModelwithExemplaryApplications
43
a price forecast for the next day is made by an agent.
The bid price p
i,h
including start-up costs in hour h is
defined as follows:
p
i,h
=
max c
var,i,d
c
s,i
t
u
if p
h
<
c
var,i,d
∧
i
∈
BL
c
var,i,d
c
s,i
t
s
if p
h
>
c
var,i,d
∧
i
∈
PL
c
var,i,d
otherwise
(3)
c
s
,i
start-up costs
t
u
number of continuous unscheduled
hours per day
t
s
number of continuous scheduled
hours per day
p
h
predicted price for hour h
M
set of all operation-ready power
plants
BL
⊂
M
set of base load power plants
PL
⊂
M
set of peak load power plants
In addition, SupplyBidders can increase the bid
price for their power plants by a mark-up value.
According to the standard economic model of
perfectly competitive markets, market prices for a
respective good are determined by marginal prices at
all times. However, in order to cover capital
expenditures and fixed costs market prices need to
rise above marginal costs of supply at least in some
periods. This reasoning is based on the peak-load
pricing concept (Boiteux, 1964). One potential
remedy is to include an additional mark-up factor in
the bid price of supply capacity, which is
implemented in the PowerACE model.
The value of the mark-up factor depends on the
relative scarcity in the market; a higher scarcity
induces a higher mark-up, which is added to the bid
price:
p
i,d
markup
=p
i,d
+ markup
h
(4)
After determining the offered volume and price
for each hour of the following day the bids are
submitted to the day-ahead market auctions. A
comprehensive and formal description of the
original short-term bidding algorithm can be found
in Genoese (2010).
3.3 Coupling of Interconnected
Markets
European electricity markets are interconnected via
high-voltage transmission lines. Since electricity
flows according to physical laws and interconnector
capacities are limited, these capacities have to be
allocated to market participants otherwise
transmission lines might get congested. Management
methods are required to avoid congestion and to
efficiently use cross-border transmission capacities.
Since 2010, a market coupling approach has been
implemented in Central Western Europe which
complies with the European Union’s general
principles of congestion management (e.g. non-
discriminatory, market-based). Market coupling
describes the implicit auctioning of interconnection
capacity through power exchanges for predefined
zones (market or bidding areas). The market
coupling operator clears the energy markets of the
participating market areas simultaneously and
determines implicitly the commercial flows between
markets areas as well as the prices. The market
coupling approach maximizes the social welfare by
optimizing the selection of bids while considering
limited transmission capacity. The transmission
capacity is determined up-front based on defined
rules (EPEX Spot, 2010).
In accordance with the CWE Market Coupling
architecture, market coupling is implemented within
PowerACE for the day-ahead market and market
participants submit their bid curves to the local
power exchanges based on the described method in
section 3.2.
In PowerACE the MarketCouplingOperator
takes over all processes related to the market
coupling. For that purpose, the operator receives all
day-ahead bids from the local power exchanges.
Market coupling itself can be formulated as an
optimization problem with the objective to
maximize social welfare. Since PowerACE currently
only considers hourly bids with a fixed price, the
original COSMOS algorithm used for the CWE
Market Coupling (APX-ENDEX et al., 2010) can be
simplified and the mathematical problem is
formulated as follows (e.g. Meeus et al., 2009):
max
q
P
b,d
Q
b,d
q
b,d
d
- P
b,s
Q
b,s
q
b,s
s
b
(5)
subject to
q
b,d
, q
b,s
≤ 1
(6)
P
b,d
Q
b,d
q
b,d
d
P
b,s
Q
b,s
q
b,s
s
+ Cap
b,b
to
b
to
Cap
b
from
,b
b
from
=
0
(7)
Cap
b(from),b(to)
≤ Cap
b(from)
,
b(to)
max
(8)
ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence
44
where the indices d and s indicate demand and
supply variables, respectively. b denotes the market
(bidding) area, P
i
the bid prices and Q
i
the bid
volumes. q
b,d
and q
b,s
are the acceptance rates of the
corresponding demand and supply bids.
Cap
b(from),b(to)
equals the determined capacity
between two market areas. Cap
b(from),b(to)
max
denotes the
upper limit for the transmission capacity between to
market areas and is given exogenously based on
current values from the European Network of
Transmission System Operators for Electricity
(ENTSO-E).
The constraints ensure that supply and buy bids
do not exceed their maximum volume (6), that
supply and demand including exports as well as
imports in market areas are balanced (7) and that the
limitation on the transmission capacity (8) is not
violated. In this form, the problem is linear and can
be solved with common solvers.
Optimization results are the acceptance rates for
each submitted bid and the commercial utilization of
transmission capacity. Furthermore, the algorithm
determines the market prices of electricity one day-
ahead of delivery in the coupled bidding areas and
the implicit prices for transmission capacities, which
are only different from zero if lines are congested.
Prices are sent to the local market areas and
processed by the supply agents.
3.4 Long-term Investment Planning
In the model generation companies can also make
decisions regarding their long-term capacity
extension through investments in new power plants.
The responsible agent is called InvestmentPlanner.
The basic methodology is based on a discounted-
cash flow valuation of predefined technology
options. For that purpose the InvestmentPlanner
makes a forecast of the expected hourly electricity
prices during the investment period and calculates
the expected yearly gross profit. After accounting
for fixed costs and capital expenditures, the net
present value is calculated. A formal description is
provided in Genoese (2010).
The quantity of the installed capacity is based on
the expected development of market shares within
the following five years taking future demand and
electricity generation from renewable energy sources
into account. As long as the net present value of the
investment options is positive and there is need for
new capacity, new power plants are built by the
InvestmentPlanner. After the construction phase,
whose length depends on the technology option, the
new power plants can generate electricity that can
then be sold in the markets.
3.5 Input Data and Technical
Implementation
For the considered market areas each thermal power
plant with a capacity of at least 10 Megawatt is
stored together with its main relevant techno-
economic characteristics (e.g. net electrical
efficiency, variable and fixed costs, yearly
availability) in the database of the model.
The model database includes investment options
for different power plant technologies with its
relevant characteristics and the electricity feed-in
from renewable energy sources. The electricity
demand is represented by the aggregated
consumption of all consumers connected to the
public power supply.
For market coupling, transmission capacities
between interconnected market areas are required.
Since not all neighbouring countries are always part
of a simulation, the electricity exchange with these
countries is based on historical values. Prices for
fuel and CO
2
emission allowances are required for
the calculation of the variable generation costs. Most
time series data is stored with hourly values, but
sometimes only less detailed values, e.g. for lignite
prices, are available.
The model’s results include the hourly spot
market prices in the simulated wholesale markets,
the investments in new capacity and the commercial
flows between interconnected market areas. Since
the model considers the wholesale day-head market
as the only trading place for electricity, bilateral day-
ahead contracts are not part of model’s results.
PowerACE is implemented in the object–
oriented programming language Java and can
simulate each hour of recent historical years as well
as future years up to 2050. The simulation runs are
comparatively quick in terms of computing time.
Yearly runs for one market area last only a few
minutes, which is a small fraction of the several
hours that optimization models with a similar
amount of details may take.
4 EXEMPLARY APPLICATIONS
The PowerACE model has been used for various
research analyses in the past. For instance, Sensfuß
et al. (2008) find a considerable impact of the
subsidised renewable electricity generation in the
short run on spot market prices in Germany. The
Agent-basedSimulationoftheGermanandFrenchWholesaleElectricityMarkets-RecentExtensionsofthePowerACE
ModelwithExemplaryApplications
45
impact of emissions trading on electricity prices is
explored by Genoese et al. (2007). The authors find
for the years under consideration that a large part but
not the totality of the CO
2
emission allowance price
is added by the generation companies to the variable
costs during the bidding process. A thorough
analysis of the model’s capacity to adequately
reproduce the main characteristics of the German
electricity market can be found, for example, in
Genoese (2010).
In the following sections, additional recent
analyses are presented.
4.1 Market Coupling between
Germany and France
Based on the algorithm described in section 3.3,
effects from a market coupling between the German
and French day-ahead electricity markets are
analysed. Both markets represent the two largest in
Europe in terms of electricity consumption and are
part of the CWE Market Coupling. To the authors’
best knowledge this is the first agent-based approach
that includes the coupling of different market areas
based on the current market situation.
The simulation of the model coupling is
performed for the year 2012. In the Single Markets
scenario, there is no coupling of the two markets, i.e.
no exchange between Germany and France is
considered. The Model Coupling scenario uses the
optimization routine for the coupling of the German
and French market areas. The electricity exchange
with other countries (e.g. between France and Spain,
Germany and Poland) is in both scenarios given
exogenously based on historical data.
The Model Coupling scenario shows lower
average prices than the Single Market scenario,
while the price decrease is stronger in France than in
Germany. The more pronounced effect for France
can be explained, to some extent, by the supply
curves’ shapes of the two market areas. The French
supply curve has only a gentle slope for a large part
of the country’s capacity because of the low variable
operating costs of nuclear power stations. However,
the small part of the remaining capacity consists of
notably more expensive fossil fuel-fired units. These
units are often called upon in the Single Markets
scenario. When coupling the markets, the expensive
units in France are less frequently used because
cheaper electricity can be imported from Germany.
The change in market prices does not imply that
all market participants, buyers and sellers, benefit.
The results in this simulation indicate that mainly
the consumers benefit from the market coupling
which is consistent with expectations given a lower
average price. The social welfare (sum of consumer
surplus, producer surplus and congestion revenue)
increases with market coupling, which could be
expected, as the clearing algorithm tries to maximize
this value.
In the Model Coupling scenario the available
transfer capacity is fully used in 65% of the cases.
The high usage of the full capacities and the price
effect of the coupling can be seen for a period of
100 hours in figure 2. Expanding (e.g. doubling) the
capacity amplifies the price reduction in both
countries; while the additional effect is smaller in
France than in Germany, the total price reduction is
still stronger in France. In case of sufficient capacity
there are identical prices in all hours, which is equal
to the situation of having one completely integrated
market.
Regarding only market coupling between two
countries, in this case Germany and France, while
the exchange with other country is based on
historical values, is, of course, a simplification.
Germany, for instance, has interconnections with
nine countries while France is connected to seven
countries. Amongst those countries are some that
take part in the market coupling as well, e.g.,
Austria, Belgium or the Netherlands. Hence, the
effects from the market coupling between Germany
and France in this paper might be overstated, since
either country would exchange electricity with other
countries, if this as well is no longer static and less
costly than the exchange with Germany or France,
respectively.
The presented results also depend on information
which is not publicly available and therefore needs
to be estimated, such as the operation and
maintenance costs of power plants. Deviations
between estimated and real world values could, of
course, alter the results of the simulation.
4.2 PowerACE LAB
Besides the computational model, there exists a
laboratory version, “PowerACE LAB”, where real-
life participants can assume the tasks of software
agents. Thus, the core agent-based simulation model
is supplemented by elements from experimental
economics and role-playing games (Genoese and
Fichtner, 2012).
In literature, two approaches are distinguished in
combining agent-based models and role playing
games. Barreteau (2001) proposes a parallel
existence of agent-based models and role playing
games. Hence, the model is rebuilt in a simplified
ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence
46
Figure 2: Simulated electricity flow and price difference (before and after market coupling) between Germany and France
for a period of 100 hours in 2012.
version as a game. The main goal of this approach is
to increase the acceptance of the model. Guyot and
Honiden (2006) develop an agent-based
participatory approach, where real participants are
integrated into the model by (partly) controlling the
agents’ actions. For this, user interfaces have to be
developed. In the PowerACE Lab version, the agent
based participatory approach is used.
Currently, in PowerACE Lab two modules exist
where human participants can interact. The
participants either simulate the power trading or the
investments in new generation capacity. In the
trading module, the participants receive the same
information as the computer agents. Each participant
has a list of daily available power plants with all the
relevant technical and economic data, e.g. installed
capacity, fuel costs and efficiency. In addition, a
forecast of the day-ahead prices is presented. Based
on this information, the participants submit their
bids. When all players have successfully completed
their task, the market clearing price is computed
analogously to the computational model. The players
have the possibility to adopt their strategies in each
round in order to maximize profits.
In the investment module, the players can carry
out investments according to the power and fuel
price forecast and by taking into account the
decreasing capacities due to the limited technical
lifetime of existing power plants.
The players’ decisions and chosen strategies can
be used to improve the behaviour of the computer
agents. Computer agents and real participants can
coexist as well in the simulations.
5 CONCLUSIONS AND
OUTLOOK
Agent-based simulation in general and the
PowerACE model in particular are useful means to
analyse different aspects of electricity markets. The
market and agent perspective as well as the
flexibility of agent-based simulation models allows
us to thoroughly analyse electricity markets and
interactions therein. The PowerACE model is a
detailed bottom-up simulation model which
integrates short-term market operations and long-
term capacity planning while the most important
market participants are represented by different
agents. The model has been successfully used for
various analyses in the context of electricity
markets.
Given the continuously changing economic and
regulatory environment in the power sector, several
enhancements to the model are currently in progress.
Agent-basedSimulationoftheGermanandFrenchWholesaleElectricityMarkets-RecentExtensionsofthePowerACE
ModelwithExemplaryApplications
47
In order to reflect the European market integration,
the model scope is extended to several market areas
which can be simultaneously run and coupled.
Model coupling clears the energy and capacity
markets simultaneously and determines an optimal
solution to the plant dispatch in the interconnected
market areas considering limited commercial
transfer capacities. The model coupling routine
presented in this paper offers a socially beneficial
opportunity to interconnect electricity markets
compared to a situation where no market coupling
occurs. The results for Germany and France show
that the average market price is lower in both
countries, while the price decrease is stronger in
France than in Germany.
The methodological approach of PowerACE has
nonetheless some limitations. Regarding the supply
of electricity, additional technical constraints
concerning the operation of power plants (e.g.
minimum downtimes or partial efficiency levels)
could further improve the model. Furthermore, the
perspective is limited to the supply of electricity,
which differs from the real world situation where
also the heat demand influences the usage of
combined heat and power plants.
Given the flexible modelling framework future
model extensions could include the development of
a generally scalable model version in order to
simulate micro-systems as well as larger systems
(e.g. Europe) with additional market elements (e.g.
intraday market). Concerning the decision making
process of agents, the refinement of the investment
module and the integration of different aspects of
uncertainty is another possibility to extend the
model. Regarding the design of electricity markets,
the remuneration of power plants by capacity
mechanisms in order to ensure system reliability is
another topic of research that is currently explored
within the model.
ACKNOWLEDGEMENTS
Recent extensions of the PowerACE model have
been partly funded by ESA². ESA² is a consortium
of universities and research institutions from five
European countries providing qualified decision
support for public and private clients in areas related
to energy and environmental policy. ESA² originated
from KIC InnoEnergy at the European Institute of
Innovation and Technology (EIT). More information
is available at www.esa2.eu.
REFERENCES
APX-ENDEX; Belpex; EPEX Spot (2010): COSMOS
description - CWE Market Coupling algorithm.
Retrieved from http://static.epexspot.com/document/2
0015.
Barreteau, O.; Bousquet, F.; Attonaty, J. (2001): Role-
Playing Games for Opening the Black Box of Multi-
Agent Systems: Method and Lessons of Its
Application to Senegal River Valley Irrigated
Systems, In: Journal of Artificial Societies and Social
Simulation 4 (2).
Boiteux, M. (1964): Marginal Cost Pricing in Practice.
London: Prentice-Hall.
EPEX Spot (ed.) (2010): Project Document – A report for
the regulators of the Central West European (CWE)
region on the final design of the market coupling
solution in the region, by the CWE MC Project.
Retrieved from http://static.epexspot.com/document/7
616/01_CWE_ATC_MC_project_documentation.pdf.
Genoese, M. (2010): Energiewirtschaftliche Analysen des
deutschen Strommarkts mit agentenbasierter
Simulation. Baden-Baden: Nomos.
Genoese, M.; Fichtner W. (2012): PowerACE LAB:
Planspiel Energiewirtschaft. In: WiSt 41 (6), pp. 335-
339.
Genoese, M.; Sensfuß, F.; Möst, D.; Rentz, O. (2007):
Agent-Based Analysis of the impact of CO
2
Emission
Trading on Spot Market Prices for Electricity in
Germany. In: Pacific Journal of Optimization 3 (3);
pp. 401-424.
Guerci, E.; Rastegar, M. A.; Cincotti, S. (2010): Agent-
based Modeling and Simulation of Competitive
Wholesale Electricity Markets. In: Handbook of
Power Systems II. Berlin, Springer, pp. 241–286.
Guyot, P.; Honiden, S. (2006): Agent-based Participatory
Simulations: Merging Multi-Agent Systems and Role-
Playing Games. In: Journal of Artificial Societies and
Social Simulation 9 (4).
Meeus, L.; Vandezande, L.; Cole, S.; Belmans, R. (2009):
Market Coupling and the importance of price
coordination between power exchanges. In: Energy
Economics 34 (3), pp. 228–234.
Sensfuß, F. (2007): Assessment of the impact of
renewable electricity generation on the German
electricity sector: An agent-based simulation approach.
Retrieved from http://digbib.ubka.uni-karlsruhe.de/
volltexte/documents/188330.
Sensfuß, F.; Genoese, M.; Ragwitz, M.; Möst, D. (2007):
Agent-based Simulation of Electricity Markets - A
Literature Review. In: Energy Studies Review 15 (2).
Sensfuß, F.; Ragwitz, M.; Genoese, M. (2008): The merit-
order effect: A detailed analysis of the price effect of
renewable electricity generation on spot market prices
in Germany. In: Energy Policy 36 (8), pp. 3086–3094.
Tesfatsion, L. (2006): Agent-Based Computational
Economics: A Constructive Approach to Economic
Theory. In: Judd, K. J.; Tesfatsion, L. (eds.).
Handbook of Computational Economics, Volume 2.
ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence
48
Agent-Based Computational Economics. North-
Holland.
Ventosa, M.; Baı
́
llo, Á.; Ramos, A.; Rivier, M. (2005):
Electricity market modeling trends. In: Energy Policy
33 (7), pp. 897–913.
Weidlich, A.; Veit, D. (2008): A critical survey of agent-
based wholesale electricity market models. In: Energy
Economics 30 (4), S. 1728–1759.
Agent-basedSimulationoftheGermanandFrenchWholesaleElectricityMarkets-RecentExtensionsofthePowerACE
ModelwithExemplaryApplications
49
