Powering Precision: Ensemble Learning for Superior Battery Health Estimation

In recent years, there has been a notable shift from fossil fuels to renewable energy sources due to energy challenges and the scarcity of oil resources. This transition has significantly increased the adoption of battery-powered electric vehicles (EVs) and hybrid electric vehicles (HEVs), replacing traditional internal combustion engine (ICE) vehicles. A crucial element of EVs and HEVs is the Battery Management System (BMS), which is essential for ensuring the battery pack operates safely, reliably, and efficiently. As batteries experience numerous charge and discharge cycles, their performance inevitably degrades, which can lead to failure and potentially catastrophic outcomes if the battery unexpectedly reaches the end of its life. This paper presents an innovative machine learning-based method to estimate battery health and mitigate related risks on real time battery data. The study employs proprietary data collected from Nickel Cobalt Aluminum (NCA) chemistry cells under dynamic driving conditions. Machine learning models, using ensemble algorithms, are trained to estimate the State of Health (SoH) of the battery pack. This approach has yielded a minimum Root Mean Square Error (RMSE) of 7.8%, indicating significant accuracy.