An improved comprehensive SOC prediction method based on adaptive particle filter

It's important for battery safe use and performance of the play to predict the capacity changes and the state of charge (SOC) of the battery accurately. However, accuracy of SOC forecast is affected by SOC definition, battery model and algorithm. In this paper, the factors which affect battery capacity are analyzed based on experiment. Storage capacity decay, recycling capacity fade and dynamic changes of the current effect are modeled, considering the influences of temperature, current ratio and so on. Based on the models, the definition of SOC was modified from the perspective of capacity. Besides, prediction algorithm is a key factor affecting the accuracy of SOC prediction. Compared with other SOC prediction methods, Particle Filter (PF) is a complete nonlinear filter. Due to the large amount of computation, the practicability of Particle Filter is limited. Thus, in order to take into account the calculation quantity and accuracy, adaptive particle filter based on Kullback-Leibler distance (KLD) sampling is introduced. Finally, the new method based on modified SOC definition and adaptive particle filter is verified by experimental data of federal urban driving schedule (FUDS) condition.