MODELING STRATEGIC AND OPERATIONAL DECISION-MAKING –
AN AGENT-BASED MODEL OF ELECTRICITY PRODUCERS

Emile Chappin, Gerard Dijkema, Koen van Dam and Zofia Lukszo
Faculty of Technology, Policy and Management
Delft University of Technology
Jaffalaan 5
2628 BX Delft, the Netherlands
E-mail: E.J.L.Chappin@TUDelft.NL



KEYWORDS
Agent-based modeling, agent architecture, ontology,
electricity producers.
ABSTRACT
Essential steps in building agent-based models is to
conceptualize, quantitatively representing and implementing
the behavior of agents. A formal approach is presented to
model the behavior of consumers, institutions or companies
as rational decision-makers. Key element is a generic
ontology for socio-technical systems and a framework for
conceptualizing actors as agents with strategic, tactic and
operational behavior. The approach was used to represent the
behavior of electricity producers who must decide on
investments, day-to-day operations and deal with CO2
emission-rights. This exercise amply demonstrates the value
of using a formal conceptual framework in the development
of agent-based models.
INTRODUCTION
Infrastructures for transportation, telecommunication, natural
gas, electricity, etc. are essential for a modern industrial
society. Infrastructure networks are technologically advanced
large scale systems, which development, structure and
operation is decided upon by a myriad of actors. These
infrastructures are socio-technical systems, which are defined
as systems in which physical elements are embedded in a
social network of actors and institutions. The electricity
infrastructure system, for example, not only consists of the
physical infrastructure for electricity production,
transportation and distribution, but also spans electricity
producers, network operators, a variety of consumers,
financial institutions, knowledge and service providers. Its
evolution and operation is determined by discrete events,
decisions by incompletely informed (ir)rational, agents on
continuously operating, interconnected systems. As a
consequence, the dynamics of this and other infrastructure
systems cannot be predicted from scientific and engineering
principles alone. Indeed, management science uses
psychology, sociology and economy.
We believe that combining knowledge from both engineering
and management science is required to successfully explore
the emergent behavior of these systems by simulation.
Appropriate models for socio-technical systems must
represent a great many external factors, apparatus,
technologies or facilities and a spectrum of actors. The latter
determine government policy, develop and interpret
regulations, invest under uncertainty in a global, continental
or regional market, respond to consumer demand and
preferences, ensure proper operation, avoid congestion and
so on.
Particularly agent-based models (ABM) appear suitable to
explore and gain insight in the dynamics of socio-technical,
systems: in a socio-technical system, a subsystem of actors
and a technical subsystem interact on different levels of
aggregation. Agent-based simulation models can cope with
discrete events, multi-level (disaggregate) decision-making,
emergent markets (Dam et al., 2007; Nikolic et al., in print)
More important, ABMs offer a good way to include the
social and behavioral aspects. An ABM model describes the
behavior of parts of the system, the system behavior emerges
from the behavior of the smaller autonomous parts (Dam and
Lukszo, 2006). Therefore, ABM simulations can be used for
explorative research with different scenarios.
There is a wide range of application domains for ABM, since
agent-based modeling originates from different research
domains, such as social sciences, mathematics and from
artificial intelligence (Wooldridge and Jennings, 1995;
Bussmann et al., 1998; Jennings, 2000). The agent paradigm
in agent-based research can be very conceptual and has to be
operationalised to be applicable in simulation.
A modular and flexible environment to set-up and execute
ABM simulations has been developed (Nikolic et al., 2004;
Nikolic et al., 2006). By using an ontology as its foundation
(Dam et al., 2006) effective representation and
implementation for a range of problems in a top-down
procedure has become feasible. The remising ‘simulation
engine’ is suitable for doing experiments in different
domains.
In this paper we focus on the conceptualization and
formalization of agent behavior with respect to strategic and
operational decisions. The case study presented addresses
part of the electricity infrastructure, notably the electricity
production sector where producers must deal with CO2
emission-rights trading. The main objective of the case study
was to obtain insight in the effect of the policy instrument
CO2 emission-trading (CET) on the types of power plants
power producers prefer to build and the power generation
portfolio that emerges from their decisions over time.

The main question addressed in this paper is: How can the
decision-making process of electricity producers,
investing in and in control of power plants, be modeled
for use in an agent-based model?
The remainder of the paper is structured as follows: First the
system description and the background of the case are given.
Second, a generic conceptual framework for socio-technical
systems consisting of agents and physical installations is
presented. Third, the decision-making process of electricity
producers is operationalised using the presented framework.
Some results are given to illustrate the advantage of the
approach used. Conclusions are given and some areas for
future work are identified.
SYSTEM DESCRIPTION
The system studied is that wherein electricity producers play
a pivotal role: the electricity infrastructure. Today, electricity
producers must effectively operate in power-exchange
markets, but also in markets for fuels, capital and emission
rights. The electricity infrastructure is a socio-technical
system: The social network is composed of power production
companies, retail companies and consumers that trade on
different markets. Governments and regulatory bodies are
also part of this network. Power generation facilities, power
grids and consumers equipment form the technical network.
Figure 1 shows the interdependencies of the social and
technical networks. Electricity producers thus must operate
and invest in power plants respecting current rules and
regulations. Consumers invest and operate their end-user
equipment. Suitable power grids operated by distribution
companies and/or controlled by government connect the two.
Each of these must anticipate and act upon demand, market
and regulatory developments expected in interdependent
social and technical subsystems.
CET affects producers decisions to invest by forcing the
possession of permits for emitting CO2 (Laurikka and
Koljonen, 2006; Olsina et al., 2006). The amount of permits
issued is limited and can be traded amongst parties on
emission markets (Svendsen and Vesterdal, 2003; Vesterdal
and Svendsen, 2004; Schleich et al., 2006). CET is
implemented because it is assumed that it would lead to the
most cost effective emission reduction due to the “invisible
hand” (Smith, 1776; Svendsen, 1999). Because the long term
impact of CET is unknown and serious experience is lacking,
we conjectured that agent-based simulations could help to
provide insights (Chappin and Dijkema, 2007a).
The main impact of CET on total sector CO2 emissions
would emerge from altered investment decisions of power
producers (Chappin and Dijkema, 2007a). Governments
implemented this instrument because it is assumed that it will
lead to a less CO2-intensive generation-portfolio. However,
the producers, the agents in the electricity infrastructures, are
autonomous. History shows that individual producers do not
exhibit the same decision-behavior. Furthermore, investment
and disinvestment decisions are discrete events about capital-
intensive pieces of equipment. Agent-based models are
suitable for explicitly simulating this. Our research into CET
effects is of an exploratory nature, mainly because of the lack
of historic data on emission-trading and its impact.
Exploratory modeling for decision-support is very well
possible with agent-based modeling.
CONCEPTUAL FRAMEWORK FOR SOCIO-
TECHNICAL SYSTEMS CONSISTING OF AGENTS
AND PHYSICAL INSTALLATIONS
The conceptual framework presented here consists of the
definition of an ontology, which is a formal specification of
concepts from a certain domain. Here the domain is that of
socio-technical systems and the ontology aims at being
generic so that different application domains can be
expressed with it. The concepts defined in the ontology form
the communication language of the agents as well as the
framework in which the decisions of the agents is formalized.
The concepts are abstract descriptions of classes.
Figure 2 shows a fragment of the ontology that describes the
agents and physical installations as nodes. Agents in this
model own physical apparatus that are on the level of
physical assets. Physical assets do not act themselves as they
are passive and controlled by agents owning them. The main
concepts from Figure 2 are introduced below:
• Agent: Representation of an actor or institution. Can be
the owner/controller of physical installations. Agents
have decision-making abilities and can interact with other
agents, forming a social network. Examples are the
electricity production companies, the world market, the
power exchange, etc.
• Technology: A physical installation that transforms
physical flows. Technologies cannot make decisions

Figure 1. Socio-technical Power Production System
(Chappin and Dijkema, 2007b)