Study On Fault Diagnosis Method Of Power Equipment Based On Deep Neural Network And Bayesian Network

The power transformer is one of the key equipment in the power supply system,and its normal operation is closely related to the safety of the whole power supply system.The study of power transformer fault diagnosis is very important for finding potential faults and improving the safety of power system.Therefore,it is necessary to analyze faults in time.Considering the complex mechanism of transformer fault,the study focuses on extracting the state characteristics from the data of dissolved gas content(DGA data)and partial discharge data(PD data)in transformer oil,and studies the transformer fault diagnosis method combined with machine learning theory.Through in-depth analysis of the structure and algorithm of deep belief networks,restricted Boltzmann machine,to overcome their shortcomings,to improve the traditional RBM using Bias regularization algorithm,which makes the traditional deep belief network is improved,in order to build a Bias regularization deep belief network model(BR-DBN).The specific training process and gives the network,on the basis of research on fault diagnosis method of power transformer.The traditional depth belief network has large scale,difficulty and long training time,which leads to the longer time of fault discrimination.Based on this,a new method of transformer fault diagnosis based on BR-DBN is proposed in this paper by using the data of dissolved gas content in power transformer oil.According to the characteristics of DGA data and several common fault types of transformers,a power transformer fault diagnosis model based on BR-DBN and DGA data is constructed,and the detailed diagnosis steps and implementation process are specified.This method reduces the complexity of the network,enables the network to keep the accuracy while speeding up the computation speed,thereby improving the convergence speed of the network,saving the training time of the network,and shortening the time of fault identification.The simulation results show that the transformer fault diagnosis method based on BR-DBN and DGA data is superior to DBN and BPNN in the diagnosis performance,which verifies the effectiveness of the method.Based on in-depth analysis of the research status of partial discharge monitoring and fault diagnosis,aiming at the existing problems,a partial discharge pattern recognition method based on BR-DBN is proposed based on partial discharge data.According to the characteristics of PD signal data and its common discharge types,a partial discharge pattern recognition model of transformer based on BR-DBN is constructed,and the detailed identification process and steps are given.In this model,the feature of partial discharge signal is extracted by the idea of layer by layer compression,and the BR-DBN method is used to recognize the type of discharge.In the simulation experiment of discharge type recognition,the results show that the partial discharge pattern recognition method based on BR-DBN has higher diagnostic accuracy and better convergence performance than DBN and BPNN methods,which verifies the effectiveness of this method.