A novel hybrid Deep Learning model for State of Charge estimation in Lithium-ion batteries for Electric Vehicles

An effective battery management system (BMS) relies on accurately estimating the state of charge (SOC) of Lithium-ion (Li-ion) batteries. Traditional methods face difficulties in accurately capturing the non-linear behaviour of batteries. This study presents a combined approach that improves the accuracy of SOC across different temperatures by using Deep Neural Network (DNN) and an Extended Kalman Filter (EKF). The DNN estimates the SOC at various temperatures, and the EKF refines these predictions by adding actual sensor measurements to increase estimation accuracy. The DNN-EKF model significantly increases SOC estimation by lowering RMSE and MAE of 94% and 95% respectively when compared with a standalone DNN model. This hybrid method offers a steady and temperature friendly way to estimate the battery’s SOC in real time for BMS.