A SOM AND EXPERT SYSTEM BASED SCHEME FOR FAULT DETECTION AND ISOLATION IN A HYDROELECTRIC POWER STATION

摘要

CARTIF has developed and implemented a fault detection and isolation system to be exploited in two hydroelectric power stations. The system has several components, but this communication focuses on the neural networks and expert system subsystems. The expert system stores the knowledge acquired by operators throughout their careers and is able to detect and isolate a great range of malfunctions and failures. On the other hand, the neural network subsystem has been trained with data automatically collected for one year. The neural network subsystem is both capable of detecting abnormalities previously known, as the expert system does, and the new cases that will occur. These systems are integrated in a predictive maintenance scheme.