A comparison of two ontologies for agent-based modelling
of energy systems
James Keirstead∗
C410 Roderic Hill Building
Dept of Chemical Engineering
Imperial College London
SW7 2AZ
London, United Kingdom
j.keirstead@imperial.ac.uk
Koen H. van Dam†
Delft University of Technology
Faculty of Technology, Policy and Management
P.O. Box 5015
2600 GA Delft
the Netherlands
k.h.vandam@tudelft.nl
ABSTRACT
Conceptualisations formalised in ontologies are useful to pro-
vide an interface between people (e.g. between modellers and
stakeholders), people and computers (e.g. data entry), and
to ensure interoperability between software elements (e.g.
communication between agents). As such, ontologies are
useful for modelling purposes: with a formal definition of
concepts, no misunderstanding about the intended mean-
ing is possible. In this paper two ontologies designed inde-
pendently for modelling applications in energy systems are
discussed. The first ontology is designed for socio-technical
infrastructure systems and it has been applied in a wide
range of domains, while the second was developed with a
focus on urban energy systems in particular. The different
motivations for the development of these ontologies are ad-
dressed, one comparable key class is examined and lessons
learned from these developments are presented.
Categories and Subject Descriptors
I.2.4 [Artificial Intelligence]: Knowledge Representation,
Formalisms and Methods—ontologies
General Terms
Standardization
Keywords
Energy systems, ontology, interoperability, agent-based mod-
elling
1. INTRODUCTION
Energy systems are an integral part of modern societies,
powering economic activity, transportation, building climate
control, lighting and many other applications. Yet their per-
vasiveness means that they are also extremely complex and
difficult to model. Disciplinary perspectives on energy sys-
tems range from macro-economic and policy issues (e.g. [13])
to the psychology of domestic consumption (e.g. [1]) or the
∗Dr. Keirstead is Research Fellow and Team Leader with
the BP Urban Energy Systems project at Imperial College
London.
†Dr. van Dam is researcher at the Energy and Industry
group of the Faculty of Technology, Policy and Management
of Delft University of Technology.
engineering specification of technologies and networks (e.g.
[5]).
As a result, energy systems models are typically devel-
oped for a single application with a specific question in mind.
This enables the boundaries of the modelling exercise to be
clearly defined, reducing uncertainties that might otherwise
impede progress. However, limiting the scope of a model in
this way creates two problems. First, the model may fail
to capture important cross-disciplinary interactions within
the energy system or, at the very least, fail to explicitly
identify where such links have been omitted. Second, spe-
cific models cannot be easily transferred to other contexts
(e.g. from electricity to gas), creating significant amounts
of repeat work. Developing tools that balance the needs of
context-specific analyses with interoperability is therefore a
research priority.
Ontologies are formalised conceptualisations [6], i.e. com-
mon models of data and concepts in a field of practice. They
are an increasingly popular way of improving the interoper-
ability of software models. By clarifying the definitions of
major concepts in a field, it is possible for multiple modellers
and models to reuse common components and to have a mu-
tual understanding of available data sets. These techniques
and technologies therefore suggest themselves as valuable
tools for the modelling of energy systems where the mean-
ings of modelled objects may have multiple meanings, de-
pending on the modelling discipline and technology.
The idea of using an ontology to describe energy systems
is not a new one. For example, Borst et al. [2] tried to de-
fine an ontology for the physical behaviour of “connected”
systems that “are able to exchange energy”, following the
ideas of General Systems Theory. Others study the use of
ontologies in linking energy systems and the built-up area in
cities, e.g. [16] who note that“without the shared perception
it would not be possible to develop adequate design method-
ology or approaches for design support that are systematic,
consistent, reusable and interoperable”. We have, however,
not found evidence for a widely-used ontology that can be
applied to describe different energy transformation technolo-
gies at different scales (e.g. a domestic boiler and a nuclear
power plant) and their use in infrastructure networks.
This paper examines two related but independently-developed
ontologies that have been applied to agent-based modelling
of energy systems. After first introducing the two approaches
in Section 2, we consider the differences in the specification

of a single representative class within each ontology in Sec-
tion 3. The concluding discussion in Section 4 examines the
degree of overlap between the specifications and highlights
the potential for developing a common ontology for energy
systems modelling.
2. TWO ENERGY SYSTEMONTOLOGIES
This section provides an overview of two ontologies for
modelling energy systems, both created using the Prote´ge´
Frames software [11]. The ontologies were developed inde-
pendently with different motivations and, of course, final
structures.
2.1 An ontology for socio-technical systems
This section presents an ontology for socio-technical sys-
tems developed in the Energy and Industry section at Delft
University of Technology. In this paper it is called the STS
ontology.
2.1.1 Background and motivation
Challenges for the development of network infrastructure
models – including electricity and gas grids, motorways or
supply chains – arise when trying to incorporate both techni-
cal and social systems within one model. Existing tools deal
with either the physical (e.g. models of industrial processes)
or the social networks (e.g. economic market models), but
these worlds have to be brought together in an integrated
modelling approach for socio-technical systems [14].
To support the development of agent-based models of socio-
technical systems, an ontology for this domain has been de-
veloped. The aim was to build an ontology not for one spe-
cific application domain (e.g. an electricity infrastructure),
but to find commonalities between applications and there-
fore to develop a modelling framework that is able to deal
with the reality of socio-technical network systems that are
interconnected across sectors.
Modellers use the ontology to formalize domain knowl-
edge, as language in the definition of behavioural rules and
as communication protocols between agents. In this way,
parts of the model can be re-used (e.g. re-using the model
of a certain technology with a different agent, or re-using
behavioural rules of one agent in another one, or even copy-
ing complete agents with their physical nodes into another
model) and models of different infrastructures can be con-
nected, even when they are developed by different modellers.
In the framework proposed in [14], the STS ontology there-
fore acts as a cornerstone, providing both a user interface
and a shared world model.
2.1.2 Structure
Since socio-technical systems, such as infrastructures, can
be viewed as networks, the main concept in the ontology
is that of a Node. Nodes are connected to one another by
Edges. The first distinction between Node classes is that of
SocialNode and PhysicalNode, following the requirement
that social and technical aspects of the system can be mod-
elled independently of each other. In Figure 1 a small frac-
tion of the STS ontology is presented. It shows, for example,
that an Agent ‘is a’ SocialNode and that it ‘has a’ Technol-
ogy. All nodes share slots (i.e. data fields) for describing,
among others, economicProperties (i.e. properties related
to the economics of the node) and physicalProperties (i.e.
properties related to the physical aspects of a node).
Node
PhysicalNode
SocialNode
Agent
Technology
is a
is a
is a
is a
has a
label primitive String
physicalProperties instances PhysicalProperty
economicProperties instances EconomicProperty
icon primitive String
caseLabels instances CaseLabel
outEdges instances Edge
inEdges instances Edge
outEdges instances PhysicalEdge
inEdges instances PhysicalEdge /Ownership
technologies instances Technology
status class Status
designProperties instances DesignProperty
possibleOperationalConfigurations instances OperationalConfiguration
currentOperationalConfiguration instance OperationalConfiguration
currentOperationalScale primitive Float
Figure 1: A fragment of the ontology for socio-technical sys-
tems, showing the relationship between different classes of
Nodes (Social and Physical) and some of their slots. Agents
and Technologies together form the socio-technical net-
work.
The key classes that make up the socio-technical network
are introduced below:
SocialNode A SocialNode is a Node capable of making
decisions about PhysicalNodes. Subclass Agent rep-
resents an actor in the system. This can be a single
person (e.g. an owner of a photovoltaic panel), a group
of people (e.g. the operations department) or a whole
organisation (e.g. the government). Moreover, Agent
has one class-specific slot that distinguishes it from its
super classes: a list of Technologies that the agent
owns, controls, maintains, etc.
PhysicalNode A PhysicalNode, on the other hand, rep-
resents an element in the physical world, such as an
engineered system. For PhysicalNodes, a process sys-
tem perspective is followed: in the node a transforma-
tion takes place between inputs and outputs. A Phys-
icalNode can either be a small unit (e.g. a battery)
or a very large system (e.g. a power plant). Subclass
Technology will be discussed in more detail in Section
3.1.
Edge SocialEdges (e.g. contracts) and PhysicalEdges (e.g.
a pipeline) between the nodes can be created, in which
money and information flows occur between the so-
cial elements in the model, and mass and energy flows
between the physical nodes. Social and technical net-
works can be formed, which interrelated become a socio-
technical network.
Additionally, the ontology contains a large set of prop-
erties (e.g. maintenance costs, volume, maximum capacity)