Hierarchical Energy Management Framework for Building Energy Systems

Strategic Insights -

Researchers in China have developed a new hierarchical optimization framework using an improved multi-agent deep reinforcement learning algorithm for multi-zone coordination control of building energy systems. This framework aims to optimize energy costs, indoor air quality, and energy consumption while ensuring safety constraint satisfaction. The proposed method decouples discrete operational mode decisions from continuous power adjustments and environmental fine-tuning, allowing for real-time decision-making and multi-energy synergy.

Key Takeaways:

  • The hierarchical control strategy achieved a 14.7% reduction in operational cost, a 12.6% reduction in energy consumption, and a 5.6% decrease in discomfort duration ratio compared to traditional control methods.
  • The proposed Safety Reward Engineering (SRE) mechanism enhances the training performance of the agents by encouraging them to satisfy constraints and correct violation actions.
  • The multi-agent deep reinforcement learning algorithm, OCMATD3, tackles the optimal control problem of building energy systems by selecting discrete operation modes and performing continuous adjustments in different zones.
  • The case study building results demonstrated the effectiveness of the proposed hierarchical optimization framework in achieving better indoor air quality and energy-savings.
  • The research highlighted the limitations of traditional control methods in handling hybrid discrete-continuous control space and constraint satisfaction.

Key Players and Organizations:

  • Xianyang Meng, Xian Jiaotong Tong Univ, Key Lab Thermo Fluid Sci & Engn, Ministry of Education, Xian 710049, People's Republic of China
  • Jiejie Liu, Jiangtao Wu, Wanbin Dou, and Zhenjun Ma, researchers involved in the study
  • Ministry of Education, provided financial support for the research
  • National Key R & D Program of China, funded the research

Statistics:

  • 14.7% reduction in operational cost
  • 12.6% reduction in energy consumption
  • 5.6% decrease in discomfort duration ratio
  • Case study building showed the effectiveness of the proposed hierarchical optimization framework
  • OCMATD3 algorithm demonstrated better training performance compared to traditional control methods

Sources:

  • Meng, X., et al. (2025). Multi-agent Deep Reinforcement Learning-based Hierarchical Energy Management for Better Indoor Air Quality and Energy-savings In Building Energy Systems. Energy Conversion and Management, 342.
  • Energy Conversion and Management. (2025). Pergamon-elsevier Science Ltd. The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, England.