Fault Diagnosis Of Main Circuit Components Of MC-WPT System Based On Attention Mechanism Neural Network

Magnetically-Coupled Wireless Power Transfer(MC-WPT)technology is a new type of energy transmission technology,which belongs to the category of Wireless Power Transfer(WPT)technology.It is through the high-frequency soft medium to realize the transmission of electric energy without electrical contact,do not need the traditional wire connection,so it has the advantages of convenience,flexibility,safety,reliability and so on.In some specific fields,WPT technology has unique advantages that cannot be provided by traditional power transmission methods,and can greatly improve the convenience,reliability and security of power supply for electrical equipment.In recent years,with the gradual promotion and application of MC-WPT technology,the demand for timely fault diagnosis of MC-WPT system has gradually emerged.However,there is no comprehensive solution for the fault diagnosis of MC-WPT system at present.The main body of the MC-WPT system is the power electronic system,and the fault of the power electronic system is mainly the fault of the electronic components.Therefore,in the fault diagnosis of the MC-WPT system,the power electronic system fault diagnosis theory that has been formed at present can be considered.Combined with the above analysis,based on the fault diagnosis technology of power electronic system,this paper studies the fault diagnosis scheme suitable for MC-WPT system,and analyzes the effectiveness of the scheme through simulation.The main work of this paper is as follows:(1)The basic principle and system architecture of the MC-WPT system are introduced,the fault diagnosis characteristics of the MC-WPT system are analyzed based on the fault diagnosis technology of the power electronic system,the possible causes and phenomena of the fault of the electronic components of the main circuit of the MC-WPT system are analyzed,and the fault set of the main circuit of the MC-WPT system is constructed.(2)This paper introduces the common fault diagnosis methods of power electronic system,and studies the fault diagnosis scheme of the main circuit of MC-WPT system based on the neural network technology according to the characteristics of MCWPT system itself.Finally,BP neural network,RBF neural network,RNN based on LSTM and RNN based on LSTM introduced attention mechanism are selected to diagnose the fault of MC-WPT system.(3)The "black box" theory was introduced as the fault feature extraction method of the MC-WPT system.The MC-WPT system was divided into multiple modules,layer by layer,and the voltage and current waveforms obtained from each observation point were compared and analyzed.Finally,four observation points of the MC-WPT system were obtained.(4)BP and RBF neural networks are used to diagnose the fault of the main circuit of MC-WPT system,and the better diagnosis effect is obtained,which verifies the validity of the neural network method for the fault diagnosis of the main circuit of MC-WPT system.(5)In order to optimize the diagnostic effect,the LSTM-based recurrent neural network and the LSTMbased recurrent neural network with attention mechanism are introduced to obtain better diagnostic effect,and the validity of the recurrent neural network and the rationality of the introduction of attention mechanism are verified.