Research On Fault Diagnosis Of Hydroelectric Sets Vibration Based On Neural Network And Evidence Theory

It is a very complex process for hydroelectric sets to vibrate. With the demand of the power supply quality improved together with the'self-service'management pattern adopted by large medium hydroelectric power plant, it is important to insure hydropower unit stably by establishing monitoring and diagnosis system, Diagnosising its fault in early stage and finding out, eliminating system's fault in time.Neural network possessed well in non-linear mapping, and D-S evidence theory has advantage over uncertainty, which are widely used in the filed of diagnosis. A new fault diagnosis method was put forward based on neural network and D-S evidence theory in the paper, which utilizes evidence theory, fused the output single network. Each network was regarded as evidence, the output of network can be fused in time domain and spatial domain in order to improve accuracy rate of diagnosis. The method proved feasible was used in diagnosis of common fault of hydropower unit. The main study in the paper is as follows:Firstly, discussing the purpose of carrying on hydropower units fault diagnosis, then, introduce state of the art and classical method in the field of fault diagnosis in hydroelectric sets. The mechanism, major failure and failure symptom of hydropower units, three methods that can identify unit's vibration fault. With the key technology of monitoring on hydropower sets together with the viberation signal-disposed and fault feature-extracted are analysised.After reviewed the fundament principle of BP net, we analyze its relative merits, and an improved BP algorithm is put forward, then a redesigned model of GA-BP algorithms is presented grounded on researching the problem of network weights optimization.Aim at the complexity of failure for hydropower generating units, and single network can't obtain a satisfactory result for the hydropower generating units, we can decompose this complex problem into many simple issues then utilize single network to diagnosis. In this paper, with vibration spectrum and amplitude sub-network, hydropower generating units fault can be diagnosised from different side. Then, fuse the output of each network by D-S evidence theory and get final result of diagnosis. The simulation results show that it is credible and the diagnosis results are improved. The validity of this method has been proved significantly.Finally, the software of diagnosis system on hydropower generating unit vibration based on neuralnetwork and D-S evidence theory has been developed by MATLAB7.0. Summarize the main work in this paper, and point out the shortage and direction of further research in the future.