Research On Fault Diagnosis Of Power Battery Based On Deep Learning

As a source of power for electric vehicles,lithium-ion battery is an important part of electric vehicles.Its performance directly affects all aspects of vehicle performance and driving safety.Therefore,diagnosing power battery fault,predicting the occurrence of power battery's failure is vital to extend battery life,improve vehicle performance and the safety of driving.This paper first studies lithium-ion battery theory and its fault,clearly expresses battery failure's level,type and reasons by doing lithium-ion battery failure analysis through the establishment of lithium-ion battery fault tree,combined with the failure mode.The discharge simulation of lithium battery is designed.According to the simulation,the discharge experiment of lithium battery is designed to obtain the discharge data of lithium battery.The experimental data are decomposed and reconstructed by wavelet packet to extract the data features.Then,a brief overview of the deep learning is made.The cyclic neural network is selected as the mathematical model for fault diagnosis of lithium battery.According to the discharge experimental data of lithium battery,a fault diagnosis model based on LSTM neural network is designed.The network parameters are modified and adjusted to get the optimal parameters of the LSTM neural network,and then the experimental simulation are compared and analyzed with support vector machine SVM and BP neural network,two algorithms which have excellent performance in fault diagnosis.The result shows the LSTM neural network has a superior ability on lithium power battery fault diagnosis.Finally,using Eclipse software and Java language to build a lithium-ion battery fault diagnosis system for electric buses.The purpose of real-time fault diagnosis,analysis and some related processing operations of lithium battery on multiple buses and ensuring the safety of buses is achieved.