State of Charge Estimation for Lithium-Ion Battery based on Artificial Neural Network

Abstract

State of charge (SOC) estimation for lithium- ion battery is an important part of battery management system (BMS). Accurate SOC estimation can extend lifespan of the batteries and ensure safety of the batteries operated in electric vehicles. In this paper, a data-driven SOC estimation method is presented. Various artificial neural networks (ANNs) with different hidden layers are trained by a data set which consists of a series of the battery voltage, current and SOC variables obtained from a dynamic discharge test. The battery SOC is then estimated by the trained ANNs. Comparisons of the SOC estimated by the ANNs and model-based method combining with extended Kalman filter (EKF) are implemented. Root mean squared error (RMSE) and mean absolute error (MAE) of the SOC estimated by the data-driven method are very close to those estimated by the model-based method. The results prove that data-driven methods can accurately estimate the SOC of lithium-ion batteries.