Protection Fault Analysis and Management

Davidson et al. [81] propose a fault diagnosis and management called PEDA, for protection engineering diagnosis agents (PEDA). PEDA has been used since 2004 by a British utility. Its objective is to automate the analysis and management of faults recorded by SCADAs and digital fault recorders (DFR). As thousands of faults can be recorded during an event such as a storm, engineers need to be supported by a system capable of extracting the most important information from the mass of collected data.

In this case, MAS technology was used as a medium for system integration. The hardware and software used to achieve fault analysis and management can indeed change rather frequently over time, and the system has to be able to accommodate such evolutions. The proposed MAS integrates several tools, including a rule – based expert system to classify and interpret data from SCADAs, and a model – based reasoning system to validate protection operations using DFR data.

The architecture of PEDA is based on FIPA’s agent management reference model presented earlier, and therefore implements both DF and AMS services. As shown in Fig. 15.28, additional agents are used for retrieving, interpreting, ana­lyzing, etc. data, each with its own role in the system. Some of these agents encapsulate systems that were already being used in the legacy system.

ACL and FIPA subscription protocols are used for inter-agent interaction and communication, together with an ontology for disturbance diagnosis. For example, all agents subscribe to IEI agents (telemetry processors), which in turn inform them when they have identified an incident. The AMS and DF enable the MAS to acknowledge when an agent joins the system: each agent autonomously searches for other agents that can provide them the information they need.

This MAS is one of the first applications tested in the power industry. The problems the authors had to face when transferring their laboratory prototype to the real system provide useful lessons (e. g., on platform choice (here JADE), standards, user interfaces, etc.) for other projects that may try to achieve similar tests. It also builds on an approach different from the ones used in traditional MAS work, as it does not primarily focus on distributed intelligence, but rather on the flexibility of the approach, transforming a legacy fault analysis and management system into a flexible and extensible tool, in which software and hardware parts can be easily changed.