Comparative Study of Embedded Energy Management Methods Based on Machine Learning for Dual-Source Electric Vehicles

In the actual context of dual-source electric vehicles (DSEVs), efficient energy management strategies (EMSs) are essential to optimize energy distribution between batteries and supercapacitors (SCs). However, achieving real-time decision-making and adaptability in resource-constrained environments remains a significant challenge. This study addresses these challenges by comparing two advanced machine learning (ML)-based EMSs: driving mode predictor-based EMS (DMP-EMS) and multiensemble learning-based EMS (MEL-EMS). Using signal hardware-in-the-loop (S-HIL) simulation and a Raspberry Pi 4 platform, the two EMSs are evaluated based on various criteria, including performance accuracy, model complexity, training and prediction times, required computing resources, scalability, and energy consumption. The findings highlight the tradeoffs between these EMSs, offering insights into designing tailored EMSs for DSEVs under varying operational demands. While MEL-EMS excels in environments where prediction accuracy is paramount since it offers superior predictive performance suitable for applications demanding high accuracy and scalability, DMP-EMS provides a more efficient and practical solution for real-time energy management in resource-constrained settings.