Parameter Identification And SOC Estimation Of Lithium Battery Base On Improved Kalman Filter Algorithm

Accurate estimation of the state of charge(SOC)for batteries can effectively improve the utilization of energy,and guarantee the safe operation of the battery pack.It is critical for the battery management system(Battery Management System,BMS).This paper focuses on the SOC estimation of lithium-ion batteries,and takes lithium-ion batteries as the experimental object.As a result,two different SOC estimation methods are developed as follows:(1)Based on the second-order RC model of batteries,PSOGA-MIGI parameter estimation algorithm is developed.The Multi-Innovation Gradient Identification(MIGI)algorithm used in the new algorithm is a new type of parameter identification algorithms for lithium-ion batteries and it parallels time series iteration and innovation sequence iteration.(2)Another new algorithm—AEKF-ESG(Adaptive Extended Kalman Filter-Extended stochastic gradient)algorithm,which couples two processes of model parameter identification and state estimation,making the results obtained from SOC estimation are iterated to the next step of model parameter identification process.In the process of model parameter identification,it is unnecessary to separate specific RC model parameters(this process requires a certain engineering mathematical foundation and calculation process).This greatly reduces the amount of calculation of the entire SOC algorithm.Compared with traditional EKF-RLS algorithm,the experimental results show that the accuracy and convergence of the AEKF-ESG algorithm are higher under any working conditions.Since the two processes of the new algorithm are combined with each other,the noise anti-interference ability is stronger during the calculation process,especially when the model parameter identification is unstable.Overall,the new algorithm is more robust to noise than the traditional algorithm.In summary,this paper makes two different schemes in model parameter identification and SOC estimation respectively.The former scheme is mainly aimed at working conditions with higher noise,while the latter can be aimed at working conditions with lower computing power requirements and smaller working conditions fluctuations.