A Rapid State Of Health Estimation For Lithium-ion Batteries Under Inconsistent Operating Conditions

The rapid and effective assessment of the health status of lithium-ion batteries is of vital significance in ensuring the safe,stable and efficient operation of electric vehicles,as well as facilitating maintenance,retirement and residual value evaluation of power batteries.Data driven methods have garnered substantial attention as they enable the evaluation of battery State of Health(SOH)by collecting charge-discharge operating data,without the exploration of complex aging mechanisms.Nevertheless,the inconsistency between the actual operating conditions and the experimental conditions of the battery,particularly in terms of temperature and discharge current rates,as well as the incompleteness and randomness of the chargedischarge processes,significantly affect the evaluation effectiveness of data-driven methods.To address the mismatch between the adaptability,speed,flexibility,and practical application requirements of existing data-driven methods in battery health state assessment,this paper proposes a solution for estimating SOH considering the inconsistency of operating conditions with a random sampled charging segment.The research is supported by Key Research and Development Program of Shaanxi Province(2019ZDLGY15-04-02)and the specific research contents are as follows:In terms of the weak adaptability,a novel SOH estimation method integrated with adaptive feature selection and extrapolation performance improvement was proposed.The statistics features were extracted from a voltage segment in charging phase.Considering the advantages of the parametric and nonparametric model,a fused model with elastic net regression and random forest regression was established.The model can realize an adaptive feature selection and assessment during the SOH estimation.Based on cross-validation experiments with batteries under inconsistent operating conditions,the improved adaptability of the proposed method is verified with a mean absolute error of 2.16%.For the shortcoming of rapidity in practical use,a rapid and muti-feature-based SOH estimation method was proposed.The additional time feature was employed based on incremental capacity analysis.Then,the Box-Cox transformation was used to enhance the correlation between muti-source features and SOH due to the distribution differences.Finally,the fused model was used for a rapid SOH estimation with a reduced charging segment.The result shows that the proposed method can achieve fast SOH estimation while ensuring estimation accuracy,with average sampling time reduced by 56.43%.According to the lack in flexibility,an SOH estimation method using random segments was proposed.To reduce the sample redundancy,an optimization model for segment screening and parameter adjustment was established.An improved multi-objective Harris hawks optimizer was used to solve the candidate Pareto solutions.Combined with the proposed fusion model,a flexible SOH estimation method using random segments was implemented under inconsistent operating conditions.The reductant segment samples for training are reduced by12.63%.And the mean absolute error of the proposed method is no more than 4.06%.