Regression Based State of Charge Estimation Model of Li-ion Battery

Automobiles are one of the main contributors to environmental pollution. Gasoline-powered vehicles emit harmful gases, including greenhouse gases, through their exhaust systems. In this context, electric vehicles have gained significant importance in the 21st century. Most modern electric vehicles are equipped with Li-ion batteries for energy storage. Exact calculation of the state of charge of Li-ion batteries is crucial for the stable functionality of electric vehicles, extending battery life, and ensuring safety and security. Accurate SoC estimation is also vital to reliable range prediction in electric vehicles. This work compares machine learning-based state-of-charge estimation methods for Li-ion batteries, including Linear Regression, Gradient Boosting Regression, Random Forest Regression, and Multilayer Perceptron Regression. The prediction accuracy is analyzed using various performance indices, namely mean absolute error, R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value, root mean square error, mean square error, and mean absolute percentage error. The results demonstrate the superior performance of these regression models, with the Multilayer Perceptron Regressor emerging as the best model for predicting the state of charge of Li-ion batteries among the four methods with R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, MAE, MAPE, RMSE and MSE values of 99.15%, 1.50%, 2.62%, 1.90% and 0.04% respectively. The results demonstrate how machine learning approaches may effectively increase the accuracy of state-of-charge prediction, opening the door to enhanced battery management and electric vehicle performance in terms of efficiency, safety, and dependability.