Scheduling the load of multiple chillers and distributing cooling capacity across multiple zones simultaneously remains challenging for large building HVAC (heating, ventilation and air-conditioning) systems. Meanwhile, the dynamic cooling load faced by HVAC systems places high demands on the generalization of energy optimization methods. To this end, a causal graph-aided reinforcement learning energy consumption optimization approach is proposed for large building HVAC systems. Firstly, an HVAC system with multiple chillers and multiple zones is analyzed, and the causal graph of the HVAC system is obtained. Secondly, a causal graph-aided network structure is designed to extract causal features between nodes in the causal graph. Causal structural information can help reinforcement learning achieve highly generalizable decisions and improve learning speed. Thirdly, an improved Soft Actor-Critic method is proposed with double experience replay mechanism and bias-based state augmentation to improve the resilience to external disturbances. Lastly, comparison experiments and ablation experiments based on three scenarios are conducted. Compared to other baseline reinforcement learning methods, the average reward of the proposed method increased by about 10%. Consequently, the proposed method achieves an average energy savings of 6% in different scenarios without sacrificing the indoor comfort. Theoretical analysis and experimental results confirm that the proposed method offers significant performance advantages in addressing dynamic supply-demand matching and energy consumption optimization for large-scale HVAC systems.
- Article type
- Year
- Co-author
The heating, ventilation, and air conditioning (HVAC) system is a major energy consumer in buildings, with the chiller—the core component of the system—playing a vital role in meeting cooling demands by carrying heat. Therefore, flexible demand-based regulation of the chiller is essential to improve building thermal comfort and reduce energy consumption. As a nonlinear, highly coupled, and dynamic system, the chiller exhibits varying system characteristics under different operating points and environmental conditions. This necessitates a control strategy capable of adapting to diverse operating scenarios.
To address the challenge of multicondition chiller regulation in HVAC systems, an intelligent regulation method based on multiple model adaptive predictive control (MMAPC) was proposed. To analyze the dynamic characteristics of chiller under different operating conditions, its working mechanism was examined using thermodynamic theory, and a chiller control model suitable for real-time operations was established. Based on the influence of environmental factors, three representative operating conditions were identified and classified. For each condition, an incremental model predictive controller was designed using the mechanism-based model. These controllers were integrated through an adaptive weighted control variable fusion approach to form the overall MMAPC strategy. To evaluate the proposed approach, a real-time experimental platform was developed, comprising an intelligent chiller regulation unit, a supervisory computer with a large display screen, and several underlying control devices. The platform supports flexible communication configuration, high-volume data processing, and the deployment of various intelligent algorithms. Comparative experiments between single MPC control and the proposed MMAPC strategy were conducted on this platform.
Experimental results showed that the proposed MMAPC approach reduced the average tracking error by 70% compared with single MPC control. Additionally, it decreased the average overshoot between different operating conditions by approximately 75% and reduced the average standard deviation of the compressor valve opening by around 91%. These results demonstrated the feasibility and effectiveness of the proposed control strategy in achieving accurate chiller outlet temperature tracking while maintaining HVAC system stability. The developed experimental platform successfully enabled real-time data acquisition, strategy computation, and command issuance, while visually displaying system status on the large screen.
The intelligent experimental platform effectively supports real-time strategy verification and provides a practical foundation for teaching and research. The MMAPC strategy demonstrates excellent performance in multicondition chiller regulation and show strong potential for solving tracking control problems in dynamic, time-varying systems. This method lays the groundwork for deploying chiller operation optimization algorithms and contributes to energy-saving and emission-reduction goals under stable HVAC operation.
京公网安备11010802044758号