BMS algorithm using AI techniques for state of charge estimation in electric vehicles

State of Charge (SoC) estimation is a significant part of Battery Management Systems (BMS) in electric vehicles and has a direct impact on safety, reliability, and driving range prediction. This study proposes a comparative evaluation of model-based and data-driven State of Charge estimation techniques under realistic driving conditions. An Extended Kalman Filter (EKF) has been employed as a model-based estimation tool, while Feed-Forward Neural Network (FFNN) and Nonlinear AutoRegressive with Exogenous Inputs (NARX) neural networks have been employed as data-driven estimation tools. The proposed State of Charge estimation tools have been implemented and tested in a MATLAB/Simulink environment with standard driving cycles under flat, uphill, downhill, and varying road slope conditions. Battery current, voltage, and temperature information has been employed as model inputs to the estimation tools. The estimation accuracy of each estimation tool has been evaluated using Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) criteria. The results provide insight into the compromises between physics-informed estimation and data-driven generalization under the influence of slope-induced current variations.