Modeling And State Estimation Of Power Lithium-ion Batteries For Electric Vehicles

With the depletion of fossil fuels and deterioration of global environment,major economies and automobile manufacturers around the world have stepped up the development of new energy vehicles to replace the traditional fossil-fueled vehicles.As the most promising candidates,electric vehicles?EVs?have considerably attracted the attention of the industry and academia.The performance of vehicle-mounted traction batteries plays a crucial role in determining the overall performance of EVs.Accurate battery states are the prerequisites to formulate appropriate energy management strategies,prolong battery cycle life and achieve safe driving operations,and therefore has strong theoretical value and practical significance.Benefiting from the advantages concerning energy density,cycle life and environmental friendliness compared with lead-acid and nickel-based batteries,lithium-ion batteries?LiBs?have been widely adopted in EVs.However,traction LiBs often work with wide temperature fluctuations and high dynamic loads,whereby the internal electrochemical reactions present strong time-varying and nonlinear characteristics.Consequently,reliable and accurate state estimation of traction batteries is very challenging.The main work of this thesis can be summarized as follows:Considering the pronounced hysteresis phenomenon of LiFePO₄ batteries?LFPBs?,a geometrical hysteresis potential model?HPM?is proposed.Thereby,conventional models and state estimation algorithms can be applicable for LFPBs.In addition,through analyzing the effects of heat generation and dissipation,a battery thermal evolution model?TEM?is established and extensively exploited in several aspects.To cope with battery time-varying characteristics,two parameter identification approaches are designed:a differential evolution?DE?algorithm based off-line approach is proposed,with which the identification process is speeded and the accuracy is improved;Utilizing battery electric and thermal behaviors,a least mean square error?LMS?filter based on-line approach is proposed,with which not only the algorithm complexity is effectively reduced,but also the parameter reliability can be guaranteed.In view of the close dependence of battery capacity on temperature and current rate,the concept of effective current is proposed,whereby an enhanced Ah method is established based on adapted Peukert equation and coulombic efficiency.Therefore,the significant SoC estimation error at high current rates and low temperatures is reduced.To tackle the mechanism defects of Ah counting method,e.g.unable to obtain the initial value and eliminate cumulative errors,a SoC determination method through dynamic voltage mapping is illustrated.In addition,since sampling cost and charge integration accuracy is contradictory,a frequency-adjustable sampling circuit,which can effectively reduce the number of sampling points while maintaining charge integration accuracy,is designed.Due to the noise interferences in realistic conditions,a SoC estimation framework based on the adaptive extended Kalman filter?AEKF?is proposed.The nonlinear characteristic between electromotive force?EMF?and SoC,as well as the performance deterioration resulting from the unchanged noise statistical properties,can be alleviated.Besides,the recursive least square algorithm with forgetting factor?RLSF?is also employed to extract battery OCV,where both electrical and thermal behaviors are utilized,so the SoC can be tracked with a favorable balance between rapidity and stability,and SoC oscillations are effectively suppressed.Aiming at rational SoAP predictions,a method is proposed under the multiple constraints concerning SoC,SoE,temperature,voltage and current.Independent SoC and SoE estimators based on the SR-CDKF algorithm are constructed to avoid the crosstalk between states,so that the estimation accuracy degradation by the decreasing model order can be alleviated;Besides,the EMF based virtual power makes the SoE estimation more reliable.Exploiting the proposed TEM,the generally ignored temperature constraint is taken into account,and the prediction time horizon is included explicitly,which avoids the overestimation of SoAP and the damages to the battery under the condition of less constraints.