Model Identification And Nonlinear State Estimation Of Lithium Battery For Electric Vehicles

An accurate battery State of Charge is closely related to the safe driving,power efficiency,charging management and biggest continue voyage course for the electric vehicles.However,lithium battery for electric vehicles is a nonlinear time-varying system,and will be affected by complicated working condition,environment temperature and battery aging state and the measurement noise of the sensor.These influence factors cause difficulties for the accurate SOC estimation.In this dissertation,the model identification and nonlinear state estimation of Li Fe PO4 and Li Mn O4 battery for electric vehicles were studied.The main research work and results are listed as follows:1)The test bench for lithium batteries was built,and the characteristic database for two types of lithium batteries was established based on the experiment.Based on the database,the basic characteristic of Li Fe PO4 and Li Mn O4 battery were compared and analyzed.2)Carry out the research for modeling and parameters identification.a)The second order RC network equivalent circuit model and model mathematical expression are established for Li Mn O4 battery.In order to solve the problem of "data saturation" in the process of model parameter identification,the dissertation proposes a fuzzy adaptive forgetting factor recursive least squares method to identify the parameters online.The proposed method improves the use efficiency of the new data,and has a faster convergence rate and estimation precision for the real value of parameters.b)The first-order hysteresis RC equivalent circuit model and model mathematical expression are established for Li Fe PO4 battery.Modeling as constant with some small perturbation,the state space model equation of the model parameters is established.In order to solve the problem of nonlinear parameters estimation,AUKF method based on the idea of covariance matching is adopted to identify the parameters online,considering the characteristics of the slow change of the battery model parameters.The parameter estimator and SOC estimator are completed independently,and the coupling between estimators is reduced.3)Carry out the research for SOC estimation based on single model.a)In order to solve the inaccuracy problem caused by uncertainty of model,the statistical information obeying chi-square distribution has been introduced to identify model uncertainty,and a novel combination algorithm of strong tracking Kalman filter and adaptive unscented Kalman filter has been developed to estimate SOC.AUKF is used in the time segments without process uncertainty,and STUKF is used in the time segments with identified process uncertainty.b)In order to solve the problem of performance deterioration caused by the unknown statistical characteristic of measurement noise,the Wavelet transform-adaptive unscented kalman filter(WT-AUKF)method is used to estimate SOC.The noise is separated from the measured terminal voltage signal,and the noise covariance is calculated online based on the absolute deviation formula of wavelet transform.Compared with the AUKF based method,the proposed method can more quickly converge to the real value.4)Carry out the research for multi-model fusion state estimation.In order to solve the problem of universality of algorithm for different types of lithium batteries and the problem of the optimal estimation for the same lithium battery in different life cycle stages,an improved AUKF method based on the orthogonal idea between residual and innovation is used to estimate SOC of the single-channel,and the fusion estimation rules are designed.The weights of each channel are calculated based on the residual statistical characteristic.In order to reduce the calculation burden,the dual scale algorithm is used.The proposed method can provide accurate and reliable SOC estimation value for different types of lithium batteries and different life cycle stages of the same lithium battery.