Research On Lithium-ion Battery States Estimation Methods In Electric Vehicle

The battery pack,which consists of a number of cells connected in series or parallel,is the only or the main power source for most Electric Vehicles(EVs)at present.Accurately obtaining the battery State-of-Charge(SOC)and State of Health(SOH)are the fundamentals of Battery Management System(BMS)to implement the energy management strategy.However,several problems still hinder accurately estimating the SOC and SOH of Lithium-ion(Li-ion)battery in EV now.Such as,poor battery modeling accuracy,high computing burden of the estimation algorithms,poor accuracy of the states estimation,high cost of collecting the features for battery degradation,.etc.In order to solve the above issues,this thesis focuses on the Li-ion battery states estimation methods in EV.The main content of the thesis is as follows:1.Research on Li-ion battery modeling methods for states estimation.Precise battery model is the prerequisites of the model based SOC estimation.Thus,the accuracy and computing complexity of four typical battery models are compared in this thesis.In addition,extensive tests of Li-ion battery under various temperatures and current rates are carried out,in order to obtain the features of the parameter variation in the Equivalent Circuit Model(ECM).Therefore,the research can guide the selection of a suitable battery modeling method for the Li-ion battery state estimation in real applications.A Li-ion battery Open Circuit Voltage(OCV)prediction algorithm based on multiple feedback correction is proposed,which aims at reducing the measurement process of the key parameter OCV in battery modeling and feature analysis.The proposed method only needs 4.2%of the measurement time of traditional method(2h),and is able to rapidly approach the OCV measurement.2.Research on Li-ion battery SOC estimation methods based on ECM.Seven nonlinear filters,frequently used for battery SOC estimation,are compared in accuracy and computing complexity through experimental test,which can guide the selection of estimation algorithm for real applications.Since the nonlinear filters generally suffer from large computation burden,a low-complexity battery SOC estimation method with dual PI controllers is proposed.The PI controller compensates the SOC estimation errors from different sources,experimental test proves that the computing time of the proposed method is only 30%of extended Kalman filter.Moreover,the proposed method can achieve less than 1%SOC estimation error.The proposed method provides the possibility of achieving accurate Li-ion battery SOC estimation in a low cost BMS.3.Research on a novel SOC estimation framework based on the fusion of data and model.Since the accuracy of the model based SOC estimation method is limited by the performance of the battery model,a novel framework for SOC estimation based on data and model is proposed.According to the proposed framework,Li-ion battery SOC estimation only needs very few historical data to establish a precise battery model for EV with complex driving cycles.In order to finalize the framework,the thesis firstly uses Least Squares Support Vector Machine(LSSVM)and Adaptive Unscented Kalman Filter(AUKF)to estimate SOC.SOC can be estimated with less than 1%mean absolute error in different driving cycles by constantly updating the battery model with a novel moving window method.Besides,a new dynamical linear model is proposed for Li-ion battery SOC estimation,in order to improve the efficiency of the estimation framework based on the fusion of data and model.Partial Least Square(PLS)is used to dynamically linearize the battery model.The matrixes in the state space equation of battery are directly changed to one single variable.Then,the linear Kalman filter can be used to estimate the Li-ion battery SOC.Experimental results prove the effectiveness of the proposed methods.4.Research on SOH estimation method with partial charging voltage profile.SOH estimation still suffers from poor online calculation accuracy and also the difficulty of collecting the degradation features of the Li-ion battery.Thus,utilizing the fixed and simple charging process of EV,a SOH estimation method directly using the partial charging voltage profile is proposed in this paper.In order to improve the estimation accuracy,grid search is used to optimize the selection of one single voltage range.In order to further enhance the flexibility of the estimation method in reality,multi-objective optimization algorithm(Non-dominated Sorting Genetic Algorithm II(NSGA-II))is used to select two voltage ranges considering the measurement length and the estimation accuracy.Three Li-ion batteries are aged under calendar ageing test for 360 days to validate the effectiveness of the proposed method.5.Research on SOH estimation based on current pulse test.An SOH estimation method based on current pulse test is proposed to reduce the cost of collecting the battery degradation features.Since the short-term current pulse is used,the features for SOH estimation are conveniently extracted from the charge and discharge process of Li-ion battery.In order to obtain a more efficient estimation model,a novel training method is proposed for establishing the Li-ion battery SOH estimation model.Genetic Algorithm(GA)is used to simultaneously optimize the two key processes during the model establishment:select the features and choose the hyperparameters in Support Vector Machine(SVM).In order to guarantee the flexibility of the proposed method in practical,this paper considers both the complexity of the current pulse test and the accuracy of the estimation by using NSGA-II to optimize the current pulse test and the hyperparameters in SVM.The non-dominated solutions from NSGA-II provide various ways to use the proposed method in EV under the complex and variable working conditions.Two Li Fe PO₄ batteries are aged under long-term cycling degradation test for 58 weeks to validate the proposed method.