Lithium-ion battery SOC estimation method with fusion improved Kalman filter algorithm

Lithium-ion battery state of charge (SOC)estimation is very essential to its charge and discharge optimum control, task planning, reliability improvement, and etc. Kalman Filter (KF) and its derivative methods widely used nowadays have the limitations such as no concrete criteria for parameter settings, poor model adaptability under varied environment conditions, and etc.To solve these problems, an SOC estimation algorithm based on noise Variance Variable Kalman Filtering (VVKF) method is proposed. The algorithm estimates and sets the noise variances most suitable for the system current state in each iteration, which solves the problem of accuracy degradation caused by setting the initial KF noise variance value according to expert experiences. The Least Squares Support Vector Machine (LS-SVM) is applied for the measurement equation of KF. Through establishing a sample library, the influence of the variation of battery types and operating conditions on the SOC estimation accuracy is eliminated. Experiment on the lithium-ion battery data set obtained from CACLE at University of Maryland was conducted; the experiment results prove the performance improvement of VVKF over KF and the effectiveness of VVKF in SOC estimation. © 2016, Science Press. All right reserved.