Powering Precision: Ensemble Learning for Superior Battery Health Estimation

Due to energy competition and scarcity of natural gas resources in recent years, fossil fuels have been significantly replaced by renewable energy sources. Because of this, battery electric vehicles (EVs) and hybrid electric vehicles (HEVs) are getting adopted instead of internal combustion engine (ICE) vehicles. The main component of electric vehicles and hybrid vehicles is the battery management system (BMS), which is necessary to ensure that the battery pack operates efficiently, reliably, and effectively. The battery should not degrade its performance by charging and discharging too much, which can lead to serious failures if the battery is left to its end of life. This paper aims to present a novel Machine learning-based battery health estimation algorithm by mitigating risks associated with real-time battery data. This study used proprietary data collected from nickel-cobalt-aluminum (NCA) chemistry battery cells in electric vehicles. Machine learning models are trained to estimate the battery pack's state of health (SoH) using ensemble learning algorithms. With this method, we can achieve maximum error margin of 7.8% for real time battery data, indicating high accuracy.