A Comparative Analysis of DNN Architectures for State of Charge Estimation Using Deep Neural Networks in Electric Vehicles

In Electric Vehicles (EVs), accurate State of Charge (SOC) estimation is necessary for the best management of Lithium Ion (Li-ion) batteries. The traditional approaches have limitations in the field of accuracy, non-linearity and computational complexity. This paper presents to overcome the constraints using Deep Neural Network (DNN) based method for SOC estimation using improved predictions. The paper investigates various DNN designs based on standardized driving cycles and operating temperatures, employing performance parameters such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Maximum Absolute Error (MAX). The results show that an ideal choice for real time Battery management Systems (BMS) is a 3-layer model with the highest accuracy by eliminating the prediction errors and maintaining the processing speed.