Research On Health Monitoring Of Lithium Battery Based On Data-Driven Method

Under the trend of traffic electrification.electric vehicles are widely promoted due to their energy-saving and environmental protection advantages.Lithium-ion batteries have become the preferred power source for electric vehicles due to their high energy density and long cycle life.With the increase of charging and discharging times of electric vehicles,the performance of lithium-ion batteries continues to decline.Accurate monitoring of battery health is needed to ensure the efficient energy supply of lithium-ion batteries for electric vehicles.Data-driven approach has the advantages of objective accuracy and good generalization performance,which can provide support for accurate monitoring of lithium-ion battery health.This paper studies the health monitoring of lithium-ion batteries based on data-driven.On the basis of constructing multiple sets of health indicators,a fusion model of lithium-ion battery state of health estimation based on Stacking strategy is proposed.The battery remaining useful life prediction model of improved least squares support vector machine combined with variational mode decomposition model and sparrow search algorithm is established.The effectiveness of the proposed model and the improved model is proved by simulation.The main research work is as follows:(1)Based on the characteristics of battery capacity degradation,new health indicators are extracted.Firstly,the working principle of lithium-ion battery is sorted out,and the relationship between battery capacity fading and electrothermal parameters is analyzed.Then,according to the analysis,the relationship between voltage,current,temperature and time in eight groups of charge and discharge stages is extracted as a health factor.Finally,Pearson correlation analysis was used to analyze the correlation between health factors and battery capacity.Eight groups of health indicators were highly correlated with battery capacity.The absolute values of the correlation coefficients were greater than 52.872%.and at least five groups were greater than 91.012%.(2)Aiming at the health indicators and capacity degradation characteristics of lithium-ion batteries,a high-precision estimation model of state of health is constructed based on the ensemble learning algorithm.Firstly,the initial learner is selected as the extreme gradient boosting tree,support vector machine regression and random forest model,the meta-learner is selected as the ridge regression model,and the fusion model is established by five-fold cross validation.Secondly,the Bayesian optimization algorithm is used to optimize the hyperparameters of the fusion model.Finally,the simulation results show that the minimum mean absolute percentage error of the proposed fusion model is 0.506%,and the maximum coefficient of determination is 0.99497.At the same time,compared with the unoptimized fusion model,the average absolute percentage error of the optimized model is reduced by 23.504%at most,which proves that the proposed model can estimate the state of health more accurately.(3)The variational mode decomposition model and sparrow search algorithm are used to improve the least squares support vector machine model,which improves the accuracy of remaining useful life prediction.Firstly,according to the degradation characteristics of battery capacity and health indicators,the signal is decomposed by variational mode decomposition.Then,the sparrow search algorithm is used to optimize the parameters of least squares support vector machine.Finally,the simulation calculation is carried out.The results show that the prediction absolute error of the proposed model is 30 cycles lower than that of the least squares support vector machine.The minimum average absolute percentage error of the proposed model can reach 0.088%,and the determination coefficient can reach 0.99876,which proves that the proposed model has better accuracy and generalization ability.