Research On State Estimation And Remaining Useful Life Prediction Of Lithium Battery

In recent years,with the "double carbon" goal,the electric vehicle industry has been developing very rapidly.The battery is the power source of electric vehicles,and the State of Charge(SOC)refers to the remaining power of the battery,which reflects the range of the vehicle;the Remaining Useful Life(RUL)is used to evaluate the remaining use value of the battery and remind the user to replace the battery in time.In this paper,the SOC estimation and RUL prediction of lithium batteries are studied in depth,and the specific research work is as follows:(1)Based on the second-order RC equivalent circuit model,the research related to parameter identification is carried out.The offline identification based on the parameter estimator toolbox improves the accuracy of the offline identification compared with the second-order exponential fitting method,but the parameter estimator toolbox method cannot be used for the online identification.The interference of noise energy in the traditional parameter identification is solved.The experimental results prove that the online recognition of model parameters based on FFRLS method has high accuracy and can realize the online update of model parameters,which ensures the real-time and accuracy of parameters.(2)The research related to SOC estimation is carried out from the perspective of estimating SOC by Cubature Kalman Filter(CKF).The SOC estimation algorithm of Adaptive Square Root Cubature Kalman Filter(ASRCKF)is proposed to solve the nonpositive nature of the error covariance matrix and noise adaption problem of the traditional CKF algorithm,and based on multiple time scales,FFRLS-ASRCKF is proposed for parameter The accuracy of SOC estimation is improved by the joint estimation of identification and charge state.The algorithm is validated under HPPC and UDDS operating conditions,and the experimental results prove that the proposed FFRLSASRCKF algorithm is more accurate,with an average absolute error of only 0.61%,and has good adaptability to different initial values of SOC and different model parameters,which substantially improves the robustness of the algorithm.(3)The research of RUL prediction algorithm based on data-driven model is carried out for the problem of remaining service life prediction of lithium batteries.A Global Search Strategy Whale Optimization Algorithm(GSWOA)is proposed.The GSWOA algorithm can find the optimal parameters of the objective function quickly and accurately,and the Gated Recurrent Unit(GRU)model is used as the objective function to realize the combination of GSWOA search algorithm and GRU prediction model.In this study,the GSWOA-GRU model is used for RUL prediction to investigate the effects of different prediction starting points and different types of batteries on RUL prediction.The battery aging data from NASA is used for validation,and the experimental results demonstrate that the algorithms based on GRU,PSO-GRU and WOA-GRU have higher accuracy,the average absolute error of battery B07 is 0.4%,the prediction starting point advance prediction accuracy is almost unchanged,and the prediction range is wider and more stable.Figure[90] table[17] reference[77]...