Comparison of SOH Prediction ML Technique for Electric Vehicles

Estimating the battery states—most notably the State of Charge (SOC) and State of Health (SOH) is one of the battery management system’s (BMS) most crucial tasks. Lithium-ion batteries’ (LIB)s’ SOH and remaining usable life (RUL), two significant parameters, are generally estimated by using battery capacity. However, it is challenging to determine a LIB’s capacity for use in electric vehicles. As a solution to this existing problem, in this study, a Machine Learning (ML) model is designed to predict the SOH of LIB in Electric vehicles. In this study, the required dataset is collected from the NASA Lithium Battery Dataset. Several methods are used to pre-process the collected dataset. Later, the health performance features are derived from the lithium battery charging graph, and the Pearson correlation coefficient is used to derive health features such as charging time at constant current, charging time at a fixed voltage, voltage change rate from 300 s to 1000 s, and 200s of voltage per cycle at a time. Once the features are extracted, dimension reduction is done using the Principal Component Analysis method. Once the data is reduced using different ML m1odels such as Support Vector Machine (SVM), Linear Regression (LR) and Random Forest (RF) algorithm the SOH of the Lithium battery of the electric vehicles is predicted. To evaluate the prediction model more methodically, error analysis using various error metrics such as Mean Square Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE) values are calculated to find the most efficient model with the minimum error value. The experimental findings confirm the presented study’s effectiveness and efficiency.