Intelligent diagnosis of defects responsible for partial discharge activity detected in power transformers

This paper describes the application of cluster analysis and classification techniques for the diagnosis of partial discharge defects present in electrical power transformers. The subsequent implementation of an agent-based, decision support system (DSS) incorporating these intelligent techniques is also discussed. Successful defect classification of empirical partial discharge data, using neural networks and rule induction, affirms the application of these techniques as a suitable means of providing reliable decision support for partial discharge defect diagnosis, particularly where expert diagnostic knowledge may be scarce or ambiguous. Through the interaction of intelligent agents the DSS considers the effectiveness and diagnostic contribution of each agent (intelligent technique) before presenting a consolidated diagnosis.