Publication

Abstract : The effective battery thermal management system in electric vehicles (EVs) requires an accurate estimation of State of Charge. The study is a full-fledged case study and comparative machine learning model that incorporates battery design, modeling and data collection to enhance SoC estimation in the different thermal conditions. Several machine learning methods were used, such as the Random Forest, Ridge Regression, Lasso Regression, Support Vector Regression (SVR) and Sequential Neural Networks, to predict the SoC by utilizing the data of the battery design and thermal models. It employs real time data to train predictive models which model the complex relationships between battery temperature, voltage, current and SoC. The strengths and limitations of each method are also identified as part of the analysis along with a comparison in detail of the method with the traditional estimation techniques. Findings indicate that this machine learning based models achieve great improvements to SoC estimation accuracy even under varying thermal conditions. The study provides a highly dynamic and solid solution in the current BTMS in EVs as part of the solution to safer, reliable, and efficient battery management systems.