Research On Key State Estimation Of Lithium-ion Batteries For Electric Vehicles

The overall performance of electric vehicles is closely related to the performance of the widely used lithium-ion power batteries and their battery management system（BMS）.BMS is crucial for protecting the batteries from unnecessary damage,improving their performance,and extending their lifespan.Accurate estimation of critical battery states such as State of Charge（SOC）,State of Health（SOH）,and Remaining Useful Life（RUL）is key to improving BMS monitoring efficiency and enhancing the overall performance of electric vehicles.This article conducts the following research on the further accurate estimation of the three key states of SOC,SOH,and RUL for lithium-ion power batteries in electric vehicles:1.The one-dimensional electrochemical model and cyclic aging model of lithium iron phosphate（Li Fe PO₄）batteries are constructed to lay the foundation for the estimation of critical states of lithium-ion batteries.Starting from the microscopic particle scale of lithium-ion battery,its one-dimensional electrochemical model is established based on the charge and matter mass conservation equations in solid-phase and liquid-phase materials,and the kinetic equations of electrochemical reactions at the contact interface between solid-phase and liquid-phase,and the temperature-sensitive relationship of the physicochemical property values of lithium-ion battery is linked through the Arrhenius equation.Secondly,the validity of the electrochemical model was verified by using the experimental test data of Li Fe PO₄ battery at the temperature of-10～50℃.On this basis,a cyclic aging model for Li Fe PO4 batteries with the growth of the solid electrolyte interphase（SEI）film at the negative electrode as the main aging mechanism was developed,and the scenario of SEI film rupture was considered.The rationality of the model has been widely verified by the industry,and the main characteristic variables characterizing the aging information of Li Fe PO₄ batteries are analyzed under the test condition of temperature 25°C.2.A SOC estimation method based on a temperature-compensated open circuit voltage（OCV）estimation model is proposed.The low-current test data covering the temperature range of-10 to 50°C are analyzed,and it is found that the accuracy of the approximate OCV curve obtained by the conventional mean value method decreases at 0°C and below.To this end,a mechanistic model is introduced to generate Coulomb OCV characteristic data,which is difficult to obtain by the conventional mean value method,by characterizing the uniformity of lithium concentration in the electrode particles,and the generated data are transformed into a temperature-compensated OCV estimation model that can be run in real time.The proposed OCV estimation model is introduced into a dual adaptive extended Kalman filtering framework based on the second-order equivalent circuit model,and the synergistic estimation of the equivalent circuit model parameters and SOC is achieved.The SOC estimation performance of the proposed OCV estimation model is compared with that of the conventional mean and improved mean OCV estimation models under the same framework through different temperatures and different initial SOC error conditions.The results show that the SOC estimation method combined with the proposed OCV estimation model has better SOC estimation performance in the range of temperature covering-10 to50°C;the root mean square error of the estimation results does not exceed 1%when the initial SOC error reaches 50%.3.A SOH estimation model using Gaussian process regression machine learning algorithm with energy features as input is proposed.Different battery aging test datasets are analyzed,three energy features applicable to the data-driven SOH estimation model are proposed,and then the SOH estimation model based on Gaussian process regression is constructed with the energy features as input.The performance of the proposed SOH estimation model is validated in a validation scenario including different cycling conditions and three types of lithium-ion batteries,and compared with the current mainstream SOH estimation models.The results show that the root mean square error,mean absolute percentage error and mean absolute error of SOH estimated by the SOH estimation model proposed in this paper are less than 2%,thus verifying the wide applicability of the proposed energy feature-based Gaussian process regression SOH estimation model.4.A feature selection method based on Pearson Correlation Coefficient-Ant Colony Optimization（PCC-ACO）fusion is proposed.The evaluation scenarios,candidate features and sample grouping methods of the feature selection method are selected.The RUL prediction algorithm is introduced into the feature selection method,the algorithmic framework of the systematic evaluation feature selection method is designed,and the feature selection method based on PCC-ACO fusion is proposed.The prediction performance of the features automatically selected by the feature selection method and the features selected in the current mainstream research are compared under the same RUL prediction algorithm.The results show that the features automatically selected by the method in this paper not only reduce the minimum number of features in the current mainstream RUL model from 6 to 5,but also reduce the root mean square error of the prediction results by 4.4%,thus verifying the effectiveness of the proposed feature selection method based on PCC-ACO fusion,and also illustrating the importance of feature selection method research for early RUL prediction based on data-driven models.importance.In summary,this paper addresses the problems in critical state estimation of lithium-ion power batteries for electric vehicles,and investigates the SOC estimation method based on the temperature-compensated OCV estimation model,the SOH estimation model that is widely applicable to different cycle conditions and battery types,and the feature selection method for early prediction of RUL of lithium-ion batteries,respectively,and compares the performance with the current mainstream research to verify the The effectiveness of the proposed method is verified by comparing the performance with the current mainstream research.The research results provide technical support for more accurate cognition of key battery operation states,improving the monitoring efficiency of BMS,further improving the end-use experience of lithium-ion batteries,and enhancing the safety of battery systems.