Faulty Diagnosis Research for Boiler Overheater Based on Fuzzy Clustering and Neural Networks

In recent years, the automation of thermal process in power plants has been developed greatly. The current stand-alone capacity of thermal power generating units increases continuously, coming with more massive and complex thermal process control system. Because of the expensive equipment of thermal power generation, the failure in thermal control system could cause enormous economic losses and major accidents, so safety and reliability of power plants has become the prerequisites for thermal power plant’s best economic performance. The research on thermal power plant control system fault diagnosis has very great practical value.In this paper, by analyzing and summarizing up the research and experience of other researchers, based on the existing multi-fault diagnosis methods, an utility boiler fault diagnosis system has been developed using fuzzy clustering and neural network.The research of this paper mainly includes the following contents:(1)The establishment of a simple knowledge base for boiler fault diagnosis has been made by collecting and collating various kinds of relevant field data.(2)The research on simulation for fuzzy c-means clustering algorithm is made. This paper proposed fuzzy c-means algorithm to analyze fault samples and pre-process input sample, in order to formatting a new set of learning samples for neural network training.(3)Build a power plant boiler fault diagnosis of the sample and the modular fuzzy neural network model, and a sample of the boiler fault simulation. BP network with a single comparison of the simulation results show that: the network training speed and accuracy improved, at the same time effectively address the complexity of the BP network used in the boiler system fault diagnosis, the existence of training slow convergence and easy to fall into local minimum points. And the network structure of multi-output, not only the occurrence of fault diagnosis, but also to determine the severity of the failure to run the scene to provide a useful reference.