LSTM-Driven State of Charge Estimation for Battery Management

This research presents a method for estimating the state of charge (SOC) of lithium-ion batteries using a neural network model. The primary challenges of this study include accurately capturing the nonlinear behavior of battery SOC under various driving conditions and temperatures. The approach involved initially cleaning the dataset and calculating the SOC values. The data is then scaled appropriately, and the neural network model was trained using diverse driving cycles and a combination of these cycles at various temperatures. The trained model was evaluated on both a single driving cycle and a mixed driving cycle, which integrated all driving conditions. The results demonstrate that the model achieves an R2 score of over 0.98 in SOC estimation, indicating its robustness and reliability across different scenarios. This high level of accuracy suggests that the neural network model worked well in real-life battery management systems, making accurate SOC predictions in a wide range of operating conditions.