Fault Diagnosis Of Power Electronic Device Based On Neural Network

Power electronic technology is the new technologies of using power electronic components to control power and the transformation.Through the using of power electronic devices,complete the high efficiency of energy conversion and control.In recent years,variable flow technology obtained the rapid development,through the processing power converter technology in national economic power consumption proportion of more and more big.In the developing world,about seventy-five percent of electricity by conversion is used of power electronic technology,estimates that in the 21 st century will be around ninety-five percent.Power electronic technology has been widely used in military equipment,industrial automation control,transportation,LED lighting,pharmaceutical production and small household electrical appliances.In the event of power electronic equipment,small electronic equipment will be damage,the transportation and industrial production,is will endanger people's life and property safety,and even cause serious injuries or disaster accident,affect the normal operation of the whole national economy.Therefore,the power electronic equipment fault detection and diagnosis is very important.I refer to the reference to the scientific research achievements of predecessors in fault diagnosis,and adopt the method of spectrum analysis and neural network combining IGBT open circuit fault diagnosis for power electronic circuits.Three-phase bridge inverter circuit,for example,using MATLAB2012 a simulation software to build diagnosis circuit simulation model,the simulation of the actual system in the operation of the IGBT produces all kinds of faults,the wavelet packet decomposition is adopted to power electronic equipment fault signal feature extraction,respectively,using the BP neural network,RBF neural network,and by the particle swarm algorithm to optimize the BP network for fault diagnosis simulation model,neural network fault diagnosis is given of the study sample data set and to train neural network,determine the power electronics inverter system fault diagnosis neural network structure and thevarious parameters.By comparing the network parameters of neural network fault diagnosis methods,the effects of optimization of the neural network weights and threshold to improve the designed neural network fault diagnosis system of learning and generalization ability.The simulation results show that using the particle swarm optimization(BP)neural network fault diagnosis for power electronic devices,can effectively overcome the shortcomings of BP network,improving the structure of the neural network design,diagnosis accuracy also have greatly improved.