Accurate Battery State of Charge Prediction using Bayesian Optimized CNN2D Deep Learning Model with Flask-Based Application

Abstract

Accurate estimation of the State of Charge (SOC) of lithium-ion batteries is essential for ensuring efficient battery management and reliable operation in Electric Vehicles (EVs). This work extends the SOC prediction framework by incorporating a CNN2D-based deep learning model optimized using Bayesian Optimization for improved hyperparameter tuning and feature extraction. The proposed model utilizes battery parameters such as voltage, current, temperature, and capacity from the Panasonic battery dataset to learn complex relationships influencing SOC behavior. Experimental results demonstrate that the CNN2D model achieves lower RMSE and maximum error compared to traditional deep learning models such as LSTM, GRU, and Bi-LSTM. Additionally, a Flaskbased web application is developed to provide an interactive platform where users can upload CSV input data and obtain SOC predictions in real time. The proposed system improves prediction accuracy, reduces computational errors, and offers a practical tool for battery health monitoring and EV energy management.