State of Charge Prediction for Electric Loader Battery Based on Extreme Learning Machine

This project focuses on predicting the State of Charge (SoC) of electric loader batteries using advanced machine learning techniques. Using a publicly available EV battery charging dataset, we implemented several regression models including LightGBM, Random Forest, CatBoost, MLP Regressor, and Gradient Boosting to forecast SoC based on input features such as voltage and current. Hyperparameter optimization was performed using RandomizedSearchCV to enhance model accuracy and generalization. To provide model interpretability, SHAP (SHapley Additive exPlanations) was utilized, allowing users to understand the contribution of each feature to the final prediction. The trained models were saved and integrated into a user-friendly Flask web application that supports secure login and registration. Users can enter voltage and current values to obtain real-time SoC predictions. The frontend was developed using HTML, CSS, and JavaScript. This system aids in efficient battery monitoring and energy management for electric loaders, promoting better operational decisions and extending battery life.

Keywords: State of Charge prediction, electric loader, EV battery, regression models, Random Forest, LightGBM, CatBoost, MLP Regressor, Gradient Boosting, RandomizedSearchCV, explainable AI, SHAP, feature importance, machine learning, battery management, energy efficiency, Flask web app.