航空混合动力系统能量管理策略研究综述

Review of Research on Energy Management Strategies for Hybrid Aircraft Propulsion System

中文关键词:  能量管理策略  深度Q网络算法  深度确定性策略梯度算法  强化学习  航空混合动力系统

英文关键词:energy management strategy  deep Q-network  deep deterministic policy gradient algorithm  reinforcement learning  hybrid aircraft propulsion system

中文摘要:

      能量管理策略作为航空混合动力系统的顶层控制，用于对混合动力系统不同动力源进行能量分流，是保证系统高效运 行的基础。详细论述了各类航空混合动力系统能量管理策略，系统总结了基于规则、优化和智能3类能量管理策略的特点和研究 现状。通过阐述强化学习原理，分析了深度Q网络算法和深度确定性策略梯度算法的奖励原理、神经网络更新原理、以及各自优 缺点及适用场景，并提出基于规则类能量管理策略对于专家经验依赖性较高等缺陷，可以通过将其与基于智能算法内部创新融合 进行缓解的措施。在此基础上，展望了能量管理策略的未来发展趋势为智能内部算法、智能与其他类型算法的融合创新等，可以 为后续航空混动系统能量管理策略研究提供一定的参考。

英文摘要:

      The energy management strategy, as the top-level control of a hybrid aircraft propulsion system, is utilized to distribute energy among different power sources. It is considered the foundation for ensuring the efficient operation of the system. The energy manage? ment strategies of various types of hybrid aircraft propulsion systems were elaborately discussed, and the characteristics and research status of three types of energy management strategies based on rules, optimization, and learning were systematically summarized. By describing the principle of reinforcement learning, the reward performance, neural network updating principle, respective advantages and disadvantages, and applicable scenarios of the deep Q-network algorithm and deep deterministic policy gradient algorithm were analyzed. It is pointed out that the deficiency of rule-based energy management strategies, which relied heavily on expert experiences, can be mitigated by integrating with learning-based approaches. On this basis, the future development trends of energy management strategies are envisioned to focus on internal innovation of learning-based algorithms and integration innovation with different types of algorithms. This can provide references for subsequent research on energy management strategies in hybrid aircraft propulsion systems.