State Of Health Estimation And Remaining Useful Life Prediction Of Lithium Ion For Electric Vehicles

As the main power source of electric vehicles,the safety and reliability of power battery has always been the focus of manufacturers and consumers.Effective monitoring and prediction of battery health state can improve the reliability of battery system and driving safety,so it is necessary to conduct relevant research on the health status of the power battery.The results of battery health state estimation and prediction are helpful for the vehicle owner to improve the use conditions of the power battery,maintain the battery in time,extend the use time,replace the battery scientifically,and ensure the safety of driving.In this paper,based on the characteristic parameters of power battery charging curve,the estimation and prediction of battery health status are studied.Starting from the internal structure and working mechanism of the battery,combining with the test data to analyze the influence of the battery operating conditions on its aging speed,and analyzes the internal deterioration factors of the battery.The charging curve characteristic parameters are used as the indirect health indexes,and the battery charging capacity as the direct health indexes.Pearson and Spearman correlation coefficients are used to measure the correlation between degradation characteristics and capacity.The results show that there is a close correlation between the constant drop of equal time,voltage rise of equal time,charge time equal voltage,time of constant voltage charge,and capacity of constant voltage charge with the battery capacity.Using principal component analysis to reduce the dimension of indirect health indexes,the first principal component was extracted as the fusion health indexes to estimate the state of health(SOH)and predict the remaining useful life(RUL).The Gaussian process regression(GPR)algorithm,which can express the uncertainty of the output results,is selected to establish the SOH estimation model based on the fusion health indexes.Respectively using the conjugate gradient method(CGM)and particle swarm optimization(PSO)training GPR models’ super-parameters,and comparing the estimation accuracy of the two models.The results show that the PSO-GPR model is more suitable for SOH estimation.Therefore,the SOH of the same type of power batteries is estimated basedon the model,and a good estimation effect is obtained.It is difficult to predict the RUL of battery online based on the capacity.So this paper proposes an indirect prediction method based on the fusion health indexes.Genetic algorithm(GA)is introduced to optimize the initial threshold and weight of Elman neural network(ENN),and then an indirect prediction model of RUL is established by using the ENN algorithm optimized by GA.Finally,the prediction effect of the model is verified based on NASA battery test data.The results show that the indirect prediction model of RUL in this paper has high prediction accuracy.