An upper ontology for power engineering applications

While it is the case that ontology is application-specific, and therefore cannot be standardized by FIPA or any other agent body, all the applications that this Working Group considers are in the power engineering domain. This has two implications:

  1. Since the various applications under study are all related to the network, there can be useful interaction between these applications. For example, a post-fault analysis system and a plant condition monitoring system can improve their respective conclusions with knowledge of the other: a fault on a particular section of the network may explain an increase in condition warnings of plant located there. Since the agent development model explicitly creates autonomous components that share information to their advantage, we are no longer developing monolithic applications, and cross-communication between the agents of separate applications is desirable and advantageous.
  2. Following from this, since the applications all operate within the power engineering domain, the spheres of reference overlap, and therefore it should be possible to standardize certain ontological components. For the example given above, both the post-fault analysis and condition monitoring application agents need to know about transformers, switchgear, and network location, meaning that the ontologies for both applications will contain these concepts. These power engineering concepts can be standardized to form an *upper ontology*: the starting-point for application-specific lower ontologies to extend from.

Much effort has been expended within the power engineering community on standards for data exchange. Specifically, the Common Information Model (CIM) for interchange between energy management systems, and IEC 61850 for intra-substation communication have gained much support for increasing openness of systems. Since these standards have been defined and are increasingly being deployed within utilities, it is sensible to base an agent ontology on one or both of these standards.

Data models such as these give a hierarchy of concepts within the field, both tangible (transformers, breakers, lines) and less tangible (voltage, power, time). However, an ontology contains not just the concepts of the domain, but also functions and predicates for describing relationships between concepts within the domain. These functions and predicates are often rather application-specific, and so are mostly left to the lower ontology to define.

Using this approach, the upper ontology defines the high level terms agents need to share information about what they see on the network, such as faults and events on specific circuits and items of plant. The lower ontology covers the detailed terms required by agents within one of the application areas, such as a particular measurement, sensor address, or intermediate stage of data processing that is only of interest to other agents within that application. The “higher up” the application an agent is, the more likely it will use upper ontology terms exclusively, while the core data processing agents will tend to use lower ontology terms. This relationship is shown in Figure 1.

The upper ontology

Here we propose an upper ontology for power engineering applications, based heavily on the Common Information Model. The applications that have employed this ontology are from the condition monitoring domain (1)(2), and therefore the ontology is slanted towards condition monitoring terms and concepts. This does not mean the ontology is only applicable for monitoring applications, merely that a different application that employs this ontology will need a slightly larger lower ontology to cover all the required terms.

Using this ontology

At its root, the ontology models the three terms required for JADE messaging: Concept, Predicate, and AgentAction. The definition of a Predicate is that it can be evaluated as true or false. A Concept is everything else, including physical items (e.g. transformers), abstract things such as voltages and temperatures, and functional terms. An AgentAction is a specific type of Concept, embodying an action an agent can carry out.

The power system terms taken from CIM are shown in Figure 2. The ObjectIdentifier concept is based on the CIM Naming class, containing slots that correspond to the Naming attributes. This ontology differs from CIM in that the PowerSystemResource concept has a slot for an ObjectIdentifier; in CIM the PowerSystemResource inherits from Naming. The Timestamp currently contains only one slot for milliseconds since 1st January 1970, a standard way of measuring computing time.

Figure 3 shows the hierarchy of the two basic types of condition monitoring information: Measurements and InterpretedData. Anything that has been measured can be represented as a Measurement, including readings taken directly from sensors and historical data from files. Everything that is derived from raw data can be represented as a type of InterpretedData, and specifically as a SummaryInterpretation or a DetailedInterpretation. The output of specific analysis techniques should be added as subtypes of one of these; common examples are shown in Figure 3.

The Measurement concept requires a supporting concept of Value, shown in Figure 4. This represents a number of primitives from CIM, and therefore produces a very large but flat hierarchy. Values that are of interest for condition monitoring, but do not exist in CIM, are included, such as PartialDischarge and DissolvedGas. Note that the PartialDischarge concept can be used to represent measurements in either pC or V, with appropriate use of the unitSymbol and unitMultiplier slots. CIM defines allowable values for unitSymbol and unitMultiplier, but this is avoided for the ontology to maximize extensibility to other types of measurement.

The hierarchy of AgentActions is shown in Figure 5. These are actions that one agent can request of another, and this tree will grow significantly as new applications are implemented. Since these actions are closely aligned with the specific application, this section is included more to direct the shape of the lower ontology than to expand over time.

The ontology’s predicates are shown in Figure 6. These are terms which can be evaluated as true or false for particular values of slots. An example is atTime with slots event and time, which can be used to confirm that a given Measurement was taken at a given time, or to ask which events occurred at a given time. As with the AgentAction hierarchy, predicates tend to be application-specific and are more the domain of a lower ontology than the upper ontology.

If you develop an extension which could be included as part of the upper ontology, please contact us at


  1. V. M. Catterson, S. E. Rudd, S. D. J. McArthur, G. Moss, “On-line Transformer Condition Monitoring through Diagnostics and Anomaly Detection”, 15th International Conference on Intelligent System Applications to Power Systems, 2009 (ISAP ’09).
  2. P. C. Baker, V. M. Catterson, S. D. J. McArthur, “Integrating an Agent-Based Wireless Sensor Network within an Existing Multi-Agent Condition Monitoring System”, 15th International Conference on Intelligent System Applications to Power Systems, 2009 (ISAP ’09).

Citing this ontology

If you find this ontology useful or informative, please cite it in one of the following ways:

  1. Cite one or more of the source publications that describe the ontology problem, the need for an upper ontology, and approaches to developing a power engineering upper ontology:
    • V. M. Catterson; E. M. Davidson; S. D. J. McArthur, “Issues in integrating existing multi-agent systems for power engineering applications”, Proc. 13th International Conference on Intelligent Systems Application to Power Systems, 2005.
    • S. D. J. McArthur; E. M. Davidson; V. M. Catterson; A. L. Dimeas; N. D. Hatziargyriou; F. Ponci; T. Funabashi, “Multi-Agent Systems for Power Engineering Applications—Part II: Technologies, Standards, and Tools for Building Multi-agent Systems”, IEEE Transactions on Power Systems, Vol. 22, No. 4, November 2007.
    • V. M. Catterson, “Engineering Robustness, Flexibility, and Accuracy into a Multi-Agent System for Transformer Condition Monitoring”, PhD thesis, University of Strathclyde, December 2006
  2. Reference this site as the source of the ontology UML:
    • V. M. Catterson, P. C. Baker, E. M. Davidson, S. D. J. McArthur, “An upper ontology for power engineering applications”, available from 2010.