Utilizing machine Learning Techniques to Predict State-of-Charge in Li-ion Batteries

As the popularity of electric propulsion using batteries rises alongside the demand for renewable energy, effective battery management and monitoring are crucial for sustainability and efficiency in electric vehicles (EVs). The battery monitoring system (BMS) employs IoT/sensor networks to estimate crucial battery metrics like remaining useful life (RUL), state-of-health (SoH), and state-of-charge (SoC), using data on current status, temperature, and voltage. Machine learning (ML) and artificial intelligence (AI) are increasingly utilized to enhance BMS accuracy, addressing challenges like real-time data processing and the accuracy of estimations. This paper investigates the effectiveness of Linear Regression and Random Forest models in estimating SoC. During the hyperparameter tuning phase, the models were optimized using the Grid Search method, and their performance was evaluated at various temperatures: 25°C, 10°C, 0°C, and -10°C. The findings indicate that the models' effectiveness enhances as the temperature increases. The Random Forest model exhibited superior performance at 25°C, achieving an R2 score of 0.99646 and an RMSE score of 0.000264. This paper not only contributes to advancing Li-ion battery monitoring system, but also empowers professionals in this field to harness machine learning capabilities effectively.