Modeling And Multi-state Joint Estimation For Lithium-ion Batteries Based On Fractional Calculus

With the emergence of problems such as environmental pollution and energy depletion,the development of new clean energy has become a social hotspot.As the energy source of new energy vehicles,the accurate monitoring of the State of charge(SOC)and State of health(SOH)of Li-ion battery can help the safe and stable operation of the vehicle,which can greatly improve the battery performance and extend the battery life.In this paper,we take single cell and battery pack as the research objects,firstly,we build the experimental platform and discuss the basic characteristics of lithium battery,then we design the parameter identification experiment and dynamic cycle condition experiment for the two experimental objects respectively.Next,we model the lithium battery,introduce the fractional-order theory to establish a typical fractionalorder second-order RC model to address the problems of the integer-order model,and propose an improved adaptive genetic algorithm to identify the parameters of the established model.To address the problems that the parameters of the typical fractionalorder model are fixed and cannot be updated adaptively with the changes of working conditions and environment,a time-varying fractional-order equivalent circuit model is established by combining polynomial fitting with segmental modeling.The correctness,accuracy and robustness of the proposed model are verified under pulse charging and discharging conditions.Finally,the accuracy and convergence speed of the improved adaptive genetic algorithm are verified;and the influence of the algorithm parameters on the parameter identification results is analyzed in depth.According to the time-varying fractional-order model,the state space equations are first established,and then the basic Fractional-Order Extended Kalman Filter(FOEKF)algorithm,the Fractional-Order Adaptive Extended Kalman Filter(FOAEKF),and Fractional Order Unscented Kalman Filter(FOUKF).The performance of each algorithm is verified by the urban cycling conditions(UDDS)of the battery pack respectively,and the effects of parameters such as memory length and computational cost are explored,and the experimental results prove that the FOUKF has the best performance in the battery pack.To further investigate the applicability of the algorithms to different battery working conditions,each algorithm is further investigated by four working conditions of a single battery while keeping the algorithm parameters unchanged.The experimental results show that the FOUKF has the best performance and good adaptability to different batteries;however,the single SOC estimation algorithm does not adapt to the changes of parameters such as capacity due to temperature and aging,resulting in an increase in the error of estimation results for different types and states of lithium batteries.To solve this problem,a joint SOC and SOH estimation algorithm is proposed.Firstly,the SOH estimation algorithm is determined,and the Multi-Innovation Extended Kalman Filter(MIEKF)is established based on the EKF by absorbing the idea of fractional order memory property and introducing the Multi-Innovation Theory.The accuracy and robustness of the algorithm are verified experimentally under UDDS conditions.The effects of parameters such as memory length,time scale,and SOC initial value of FOUKF-MIEKF are then investigated in depth.The adaptability of the FOUKF-MIEKF algorithm to different operating conditions is further illustrated by experiments with a single cell under various operating conditions.Finally,the dynamic stress test conditions are analyzed in comparison with a single SOC algorithm and a joint estimation algorithm under a typical fractional-order model.The experimental results demonstrate that the time-varying fractional-order model can adaptively match the model parameters at different temperatures,and that the FOUKF-MIEKF has good adaptability to different cells,operating conditions,and temperatures,substantially improving the accuracy and robustness of the algorithm.Figure[121] table[20] reference[61]...