Scholarlink

Hybrid Electric Vehicles (HEVs), Fuel Cell Hybrid Electric Vehicles (FCHEVs) and Battery Electric Vehicles (BEVs) are key solutions for sustainable transportation, requiring advanced Energy Management Systems (EMS) to enhance efficiency and system longevity. Deep Reinforcement Learning (DRL)-based EMS offers an adaptive, data-driven approach for real-time energy optimization, yet challenges remain in designing robust reward functions that balance energy efficiency, powertrain durability, and operational stability. This chapter reviews reward function design and normalization strategies in DRL-based EMS, highlighting their impact on learning performance and system efficiency. Additionally, it examines state-of-the-art hardware testing methodologies, including simulation frameworks, Model-in-the-Loop (MIL), Hardware-in-the-Loop (HIL), and real-world implementations, enabling the transition from conceptual developments to real-world implementation. Furthermore, this chapter discusses key challenges in RL design and testing methodologies, explores potential solutions, and outlines future research directions, contributing to the advancement of intelligent energy management framworks for HEVs.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Reinforcement Learning in Energy Management System for Fuel Cell Hybrid Vehicles: A Review on Reward Design and Testing Framework

  • 章节

  • 发表日期：

    2026年

查看全文
  • Ali Sayah,
  • Marwa Ben Saïd-Romdhane,
  • Sondes Skander-Mustapha

摘要

Hybrid Electric Vehicles (HEVs), Fuel Cell Hybrid Electric Vehicles (FCHEVs) and Battery Electric Vehicles (BEVs) are key solutions for sustainable transportation, requiring advanced Energy Management Systems (EMS) to enhance efficiency and system longevity. Deep Reinforcement Learning (DRL)-based EMS offers an adaptive, data-driven approach for real-time energy optimization, yet challenges remain in designing robust reward functions that balance energy efficiency, powertrain durability, and operational stability. This chapter reviews reward function design and normalization strategies in DRL-based EMS, highlighting their impact on learning performance and system efficiency. Additionally, it examines state-of-the-art hardware testing methodologies, including simulation frameworks, Model-in-the-Loop (MIL), Hardware-in-the-Loop (HIL), and real-world implementations, enabling the transition from conceptual developments to real-world implementation. Furthermore, this chapter discusses key challenges in RL design and testing methodologies, explores potential solutions, and outlines future research directions, contributing to the advancement of intelligent energy management framworks for HEVs.