Long-Short-term Memory Neural Network to Predict State of Charge (SOC)

The energy market trend is shifting from petroleum and coal-based fuels to cleaner fuels such as Hydrogen and renewable energy sources. The rise of electric vehicle technology (EVs) and Hybrid electric vehicles (HEVs) can be seen. Hence an efficient and better battery management system (BMS) is required. State of charge (SOC) estimation is a major part of the BMS. SOC indicates the battery charge level as a percentage. It is like a fuel gauge for a battery. The estimation of SOC is done using a neural network (LSTM-NN) that has long short-term memory in this research paper. The paper also explains about tuning of hyperparameter for better performance of the network. The output is measured in terms of error instead of accuracy. Root-mean squared error (RMSE), Mean absolute error (MAE), and MAX error are used as a measure of errors.