An intelligent hybrid technique for fault detection and condition monitoring of a thermal power plant

Abstract This study presents an application of intelligent fault detection system for recognizing abnormal conditions during transient operation of a steam generator unit. Unobserved dynamics of evaporator section have been caused multiple false alarms and boiler emergency shut-downs. In order to detect faulty conditions, four different classifier agents were employed in parallel. The experimental data from real system performances during system trips were collected to train and validate the intelligent classifiers. The outlet results of all classifiers were combined using Yager's rule of fusion in order to improve the reliability and accuracy of fault detection process. The performances of the proposed fault detection system were evaluated during the unit's load variations and at different scenarios as one or two classifier(s) failed to detect the correct situations. The obtained results indicated the capability and feasibility of the proposed technique in preventing from raising false alarms by early detection of abnormal conditions.