State-of-Health Prediction for Li-ion Batteries for Efficient Battery Management System Using Hybrid Machine Learning Model

Since Lithium-ion (Li-ion) batteries are frequently used for real-time applications, evaluating their State of Health (SoH) is crucial to guarantee their effectiveness and safety. Model-based methods with SoH prediction are helpful. However, the issues with battery modelling have led to a greater dependence on machine learning (ML). As a significant step in assessing the effectiveness of ML techniques, data preprocessing has also drawn much attention. In this work, a new preprocessing method using relative State of Charge (SoC) is proposed; further, this paper describes a hybrid learning model (HLM) that combines auto-regressive integrated moving average (ARIMA), gated recurrent unit (GRU) and convolutional neural network (CNN). Data: proposed HLM uses time-series and SoC domain data; the ARIMA + GRU algorithm trains the time-series data, while CNN trains the SoC domain data. Both outputs are mean averaged to get the final output prediction. The proposed HLM is evaluated for root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) using the National Aeronautics and Space Administration (NASA’s) randomized battery usage data set (RBUDS). The results indicate that the recommended HLM is more accurate and has a smaller error margin than existing ML models.