Fuzzy Expert Systems for the Diagnosis of Component and Sensor Faults in Complex Energy Systems

Abstract

Locating the causes of malfunctions in complex energy systems is an extremely difficult task, since more than one fault mode may produce similar and possibly undistinguishable patterns of effects. This paper shows how fuzzy expert systems can exploit the available measurements from the data acquisition system to identify different component and sensor fault modes. Real sensor data (mass flow rates, pressures, temperatures, and key operating parameters) are compared with the expected values of the same quantities that are calculated using numerical models of local subsystems. This comparison simply determines if the differences between measured and expected values are “negative,” “zero,” or “positive” in fuzzy logic terms. The final objective is to verify the existence of some patterns of these attributes that univocally identify the considered fault modes. These patterns are then implemented as the set of rules forming the knowledge base of a fuzzy expert system. The proposed diagnostic methodology is tested on the gas section of a real combined-cycle cogeneration plant, and the effect of measurement noise is also discussed.