A Hybrid CNN-LSTM for Discharge Capacity Estimation of Lithium-ion Batteries

Predicting discharge capacities of Lithium-ion batteries (LIBs) is essential for safe operation of the battery in Electric Vehicles (EVs). In this paper, a Convolutional Neural Network-Long Short Term Memory (CNN-LSTM) based deep learning is proposed to estimate the discharge capacity of LIBs. The parameters such as the voltage, current, temperature and charge/discharge capacity are recorded from a Battery Management System (BMS) at various stages of the charge-discharge cycles. Data was recorded keeping the stress constant because this parameter couldn't be controlled. Two different sets of data were obtained at two magnitudes of stress values. The experiments conducted to collect the data was recorded in cycles, where each cycle was divided into 7 steps. Each testing cycle comprises of charging, discharging, rest and cross validation test. The initial layers are convolutional layers that helps in feature extraction followed by a Long Short Term Memory (LSTM) layer. The evaluation model was done using multiple train test split method. The lower values of weighted mean squared error (MSE) obtained suggests that discharge capacity estimation using CNN-LSTM is a reliable method when compared to the conventional voltage-based method. The CNN-LSTM program can further be compiled in BMS in EVs to obtain real time status for State of Charge (SOC) and State of Health (SOH) values.