Research On The Multi-scale Estimation Of The Lithium-ion Battery Capacity And SOC For Electric Vehicles

SOC estimation is a basic function of electric vehicle battery management system,and the accurate SOC not only is beneficial to improve energy utilization of battery,ensure the security of battery,and reduce cost,but also make contribution to optimize the vehicle energy management,eliminate‘the range anxiety'and so on.But for a nonlinear and several variable battery system,it is extremely difficult to obtain the accurate SOC because of the unknown initial value,the inaccurate capacity,and the noise interference.For the purpose of the exact SOC and capacity estimate,some work has been done in this paper and been summarized as follows:(1)To achieve precise modeling of lithium-ion battery,select six kinds of common equivalent circuit model,combine the feature extraction method with the optimal fitting method to accomplish the off-line parameter identification,at the same time,use the recursive least squares method with forgetting factor and joint estimation method of state and parameter to achieve the on-line parameter identification of models with hysteresis and models without hysteresis respectively.Through comparing,evaluate the performance of the above equivalent circuit models and parameter identification methods,further select1-order RC model as the basis of later estimation algorithms.(2)For the purpose of accurate SOC estimation of lithium-ion battery,instead of the traditional extended Kalman filter,use H? filter to improve the robustness of the original algorithm.Considering that the precision of off-line parameters is relatively low,realize the joint estimation of state and parameter by the means of dual H? filters,which can improve the estimation performance in great degree.Then,achieve the noise-adaptive updating of dual H? filters to improve the stability and robustness in the practical application.At the same time,use four evaluation indexes,which are the prediction accuracy of terminal voltage,estimation precision of SOC,bound error of SOC and the convergence rate with inaccurate initial value,to make a more comprehensive evaluation to the above algorithms.(3)Based on the above dual H?filters,view capacity as battery parameter and use Capacity-SOC-OCV(Open Circuit Voltage)3d surface instead of the single SOC-OCV curve,thus realize the joint estimation of SOC and capacity.According to that slowly time-varying parameter and fast time-varying,further put forward the multi-timescale H? filters.From five aspects include the prediction accuracy of terminal voltage,estimation precision of SOC,the convergence rate of SOC,estimation precision of capacity and the convergence rate of capacity,verify that this approach has a better estimation performance than the single-timescale approach.(4)In order to analyze the influence of external interference to the above design algorithms and improve the accuracy and reality of simulation process,introduce power battery to the simulation loop and set up the battery-in-loop simulation test platform,which verify the feasibility and robustness of dual H? filters algorithms.This paper devotes to achieve the real-time and accurate estimate of battery SOC and capacity,by the contrast of six kinds of equivalent circuit models with online or offline parameter identification methods,choose 1-order RC model as research object,use dual H? filters algorithms to realize joint estimation of state and parameter,considering the influence of capacity to SOC estimation,further build the joint estimation strategy of SOC and capacity based on the multi-timescale H? filters,and verify the practical application value of the above approachs by hardware-in-loop simulation test finally.