Detection of Boiler Tube Leakage Fault in a Thermal Power Plant Using K-means Algorithm based on Auto-Associative Neural Network

Abstract

The fault detection system using K-means algorithm based on Auto-Associative Neural Network (AANN) is proposed for boiler tube leakage in a thermal power plant. The normal operation state of the power plant is modeled using the AANN proposed by Kramer among various neural network techniques. The difference between the normal operation state estimation value which is the output of the model and the actual value of the main variables related to the fault is called residual. Using the residuals and residual variation of each variable, the fault detection system of boiler tube leakage is implemented. Finally, the actual fault cases of the boiler tube leakage are applied to verify the possibility of fault detection.