Optimizing Battery Charge Prediction Accuracy Utilizing Machine Learning Methods

Energy storage systems are more cost-effective when they correctly manage the capacity for lithium-ion batteries (LiBs), especially when they are used on a big scale. The design saves money, in the long run, to repair or fix LiBs less often. To determine the amount that LiBs were capable of holding, adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost), gradient boosting, light gradient boosting machine (LightGBM), category boosting (CatBoost), as well as ensemble learning models are utilized. Employing the mean absolute error (MAE), and the mean squared error (MSE) along R2 numbers, the researcher compared the accuracy with which each model could predict future outcomes. For example, the LightGBM model had the least MAE (0.102) as well as MSE (0.018) values, as well as the greatest R-squared (0.886) value, which means that its predictions were most closely related to reality. It was about the same in terms of speed among the gradient boosting as well as XGBoost models, which came next to LightGBM. The ensemble model's efficiency suggests that integrating many models might result in an overall increase in performance. In addition, the research uses Shapley additive explanations (SHAP) values to analyze important aspects influencing model predictions within the context of explainable artificial intelligence (XAI). This study found that discharge capacity is strongly influenced by temperature, cycle index, voltage, and power. This study demonstrates that Machine Learning (ML) methods can improve energy storage systems and regulate LiB in XAI.