Deep Neural Network-Based State-of-Charge Estimation of Li-ion Batteries for Electrified Transportation

Electric vehicle (EV) battery management system (BMS) depends heavily on the accurate state of charge (SoC) statistic. It impacts the lifespan, effectiveness, and credibility of the Li-ion battery (LIB). However, erratic battery dynamics and SOC fluctuations imposed by external variables like ambient temperature, battery age, and number of charge/discharge cycles make precise SoC estimation difficult. Consequently, using cutting-edge techniques is crucial for accurate SoC estimation. Deep learning models' computational intelligence and self-learning abilities have proven them beneficial for SoC estimation. This paper presents a deep neural network (DNN) based SoC estimator design approach. The model is trained and validated using the Unified Driving Schedule (LA92) drive cycle, and a comprehensive analysis has been carried out at different temperatures. Finally, the model's accuracy is calculated using various error metrics, and the results indicate that the model succeeds in achieving 0.58%, 0.69%, and 0.6% mean absolute error (MAE) at 0°C, 25°C, and 40°C respectively.