Research On Power Battery Fault Diagnosis Based On Neural Network

As the power source of new energy road sweeper,the safety of power battery has been widely concerned.In daily use,power batteries often face overcharge,over discharge and other faults,which lead to the decline of battery life.The complex operating environment may lead to uneven charging and discharging of a single battery,which may lead to safety accidents.In view of the above problems,the fault diagnosis algorithm is studied for the power battery of the new energy road sweeper,and the fault diagnosis strategy is formulated.Firstly,this thesis analyzes the principle and fault of power battery,summarizes the main fault types,and determines the fault types studied in this thesis by combining the severity of different faults.By building a power battery test platform,the battery charge and discharge data are obtained.Secondly,the battery model is analyzed,the second-order Thevenin equivalent circuit model is designed,and the full rank decomposition least square method is used for parameter identification.The accuracy of the model was verified by Simulink simulation combined with the test platform.Thirdly,aiming at the problem that the power battery cannot diagnose the fault in real time and the accuracy is low during operation,a battery fault diagnosis algorithm based on Box-Cox and LSTM was proposed,and the fault diagnosis of lithium iron phosphate battery pack was studied.Box-Cox transformation was performed on different input data to strengthen the correlation of the data.Then,deep learning was used to establish the LSTM fault diagnosis model,and iterative optimization was carried out through LSTM.Comparing the accuracy of different models,the data verification results show that the fault diagnosis method based on LSTM and Box-Cox transformation has higher accuracy for power battery pack fault diagnosis.Finally,the established fault diagnosis algorithm is applied to the test platform to verify the accuracy of the algorithm for fault diagnosis.