Research On State Monitoring And Fault Warning Of Electric Vehicle Charging Process Based On Deep Learning

With the continuous growth of electric vehicle ownership,spontaneous combustion and fire accidents of electric vehicle continue to occur,causing serious economic losses to vehicle owners and operators of charging facilities,and the charging safety problem has become a stumbling block to hinder the growth of electric vehicle and their associated industries.To solve this problem,this paper proposes an electric vehicle charging process state monitoring and fault warning method based on deep learning at the charging side in accordance with the charging features of electric vehicle.The method can monitor various physical data of electric vehicle charging,realize charging fault warning of electric vehicle,and effectively avoid false alarms caused by wrong charging data,which is very important to vigorously boost the growth of electric vehicle and their associated industries.Firstly,the research background and significance of electric vehicle charging status monitoring and fault warning are explained,and the status of its research at home and abroad is summarized;the power battery of electric vehicle,charging way,charging methods,obtainable and monitored electric vehicle charging parameters and factors that affect the safety of electric vehicle charging process are analyzed,the possible charging faults on the charging side are listed,and the fault warning idea and warning process are proposed;the charging data pre-processing method is introduced,and the criteria for evaluating the good or bad training of the model are proposed.Secondly,the principle and structure of existing deep learning methods are deeply analyzed,on which the network optimization methods of deep learning methods are summarized;the fault warning method of sliding window is proposed,which lays the foundation for judging the electric vehicle charging fault warning.Next,the temperature prediction model of electric vehicle charging process was constructed using gated recurrent unit（GRU）and the multi-parameter prediction model of electric vehicle charging process using bidirectional gated recurrent unit（BiGRU）were constructed,and the forecasting results of the models were analyzed;the temperature fault and multi-parameter fault warning thresholds were determined according to the sliding window analysis method;it was verified through experiments that the GRU model and BiGRU model combined with the fault warning method can realize the early warning of faults in the charging process of electric vehicle.Then,to improve the accuracy of electric vehicle charging process parameter prediction effectively,CNN-BiGRU hybrid neural network model is proposed in combination with the ability of convolutional neural network（CNN）to extract deep features;a multi-parameter prediction model of electric vehicle charging process based on CNN-BiGRU network is constructed,and the prediction and fault warning effects of the hybrid CNN-BiGRU model are analyzed.Finally,the results of the state monitoring and fault warning method of electric vehicle charging process studied in this paper are summarized and the main directions of future research are foreseen.