In the process of traditional battery state of charge (SOC) estimation, the extended Kalman (EKF) and unscented Kalman (UKF) methods commonly used are only suitable for linear system and Gaussian environment, although the basic particle filter (PF) algorithm can be applied to non-linear and non-Gaussian systems, there is particle degradation in PF algorithm when updating the filter, which will make the weight of sampling particles concentrate on a few particles, and make the particle set unable to truly represent the actual posterior probability distribution, which will reduce the estimation accuracy. In this paper, the improved particle filter methods are used to estimate the SOC of the battery, which can improve the estimation accuracy caused by particle weight degradation. This method is based on Thevenin battery model, the least square method with forgetting factor is used to identify the model parameters, combined with the improved particle filter algorithms, the battery SOC is estimated. The experimental results show that the accuracy of particle filter method with UKF as the posterior probability density function is higher than that with EKF as the posterior probability function, and the effect of restraining particle weight degradation is the most obvious. |