Multidimensional minimization training algorithms for steam boiler high temperature superheater trip using artificial intelligence monitoring system

Steam Boilers are important equipment in power plants and the boiler trips may lead to the entire plant shutdown. To maintain performance in normal and safe operation conditions, detecting of the possible boiler trips in critical time is crucial. As a potential solution to these problems, an artificial intelligent monitoring system specialized in boiler high temperature superheater trip has been developed in the present paper. The Broyden Fletcher Goldfarb Shanno Quasi-Newton (BFGS Quasi Newton) and Levenberg-Marquardt (LM) have been adopted as training algorithms for the developed system. Real site data was captured from MNJ coal-fired thermal power plant-Malaysia. Among three power units in the plant, the boiler high temperature superheater of unit one was considered. An integrated plant data preparation framework for boiler high temperature superheater trip with related operational variables, have been proposed for the training and validation of the developed system. Both one-hidden-layer and two-hidden-layers network architectures are explored using neural network with trial and error approach. The obtained results were analyzed based on the Root Mean Square Error for developed intelligent monitoring system.