The capacity estimation of Li–Ion battery using ML-based hybrid model

Accurate estimation of State of Charge (SoC) and battery capacity estimation is critical for optimizing the performance and reliability of lithium–ion batteries in electric vehicles and other battery-powered systems. However, challenges such as battery aging, variable driving profiles, and inaccurate SoC estimation hinder effective battery utilization. To address these challenges, we propose a novel hybrid model based on machine learning and improved coulomb counting method (CCM) that integrates real-time sensor data with a pre-trained dataset to improve SoC and battery capacity estimation. The proposed method incorporates voltage measurements along with CCM to accurately estimate SoC using real-time battery parameters. Additionally, a linear regression algorithm is employed to compare and analyze the pre-defined and real-time data, while L2 regularization is adopted to prevent overfitting in the scatter plot curve analysis. The proposed hybrid model effectively reduces cumulative errors in SoC estimation by integrating the coulomb counting method with machine learning techniques and real-time data. Based on the proposed methodology, the mean square error (MSE) obtained is 2.1 x \(10^{-4}\) ${10}^{-4}$ which is significantly lesser than the works in literature.