Battery SOC estimation based on multivariate adaptive regression splines

The voltage, current and temperature are used as input variables in this paper. The SOC estimation of the lithium iron phosphate battery is achieved by the method of multivariate adaptive regression splines (MARS). The data obtained by the test is standardized, which can be then used in the train of the SOC estimation mathematical model. After that the model is verified. The simulation results show that the proposed method can improve the accuracy of SOC estimation.

1. Introduction

Battery state of charge (SOC) which reflects the use of the battery and improves the safety of vehicle operation is one of the key parameters of battery. But it is difficult to estimate the SOC during the operation of the electric vehicle (EV) since the current fluctuation will change the temperature of the battery and it may be warmed. Therefore to choose proper method for SOC estimation is one of the main functions and research focuses of the battery management system (BMS).