Research On SOH Estimation Method Of Vehicle Lithium-ion Battery Based On Data Processing And Neural Networ

The rapid development of the global economy has caused a certain degree of energy crisis and environmental pollution and other problems,in order to alleviate these contradictions,new energy electric vehicles are favored by people as the main development direction.Lithium-ion batteries are widely used as the power source of electric vehicles,and their state of health(SOH)is of great significance for the safety and stability of electric vehicles during normal operation.Aiming at the problems of traditional SOH estimation methods that cannot accurately estimate the recovery capacity,poor feature validity,and lack of estimation accuracy,this paper proposes a lithium-ion battery SOH estimation model that takes into account the battery capacity recovery phenomenon.The working principle and main performance parameters of lithium-ion batteries are analyzed firstly,the internal and external factors affecting the decline of SOH of the battery are discussed.Next,the phenomenon of capacity decay and capacity recovery of the battery are further analyzed,and the capacity is decomposed into smooth component and fluctuation component.Then,the characteristic parameters of battery capacity attenuation are analyzed,and a filtering method combining median absolute deviation(MAD)filtering and Savitzky-Golay(SG)filtering is proposed to process the data,which improves the effectiveness of the features.Father,based on the aging law of voltage,current and temperature in charging stage,six features are put forward and decomposed into features which are highly related to smooth component and fluctuation component respectively.After feature decomposition,invalid information can be removed and data redundancy and model calculation can be reduced.Secondly,an estimation model based on improved sparrow search algorithm and Elman neural network(SSA-Elman)is constructed to estimate the smooth components.In addition,the incremental updating mechanism(IUM)is added to the model to improve the convergence accuracy and generalization ability of Elman.At the same time,an estimation model combining attention mechanism with Bi LSTM(ABi LSTM)is constructed to estimate the fluctuation components.In order to prevent over-fitting,Dropout layer and L2 regularization are introduced into the model.Next,three sets of simulation experiments are carried out based on NASA and CALCE battery datasets to verify the performance of the proposed model.The results show that the proposed method can accurately estimate SOH,with the root mean square error(RMSE)as low as 0.0054 and the mean absolute percentage error(MAPE)as low as 0.47%,which achieves better results than the direct estimation of SOH by the model.Finally,the estimation performance of the proposed model is verified on the self-built experimental platform,and the SOH estimation error is generally kept within 0.015 Ah,and the maximum error does not exceed 0.0226 Ah,which proves that the proposed method can achieve accurate SOH estimation and has high practical value.