A general multi agent-based distributed framework for optimal control of building HVAC systems

Abstract

Distributed control is recognized as an effective means for improving energy efficiency of building heating, ventilation, and air conditioning (HVAC) systems. However, existing distributed control methods ignore the physical correlations between components and zones, resulting in heavy computation load for system-level control. This paper proposes a hierarchical architecture with multi layers to describe the tree-like structure of various HVAC systems. It divides the conventional big distributed optimization task into several local optimization tasks, according to the hierarchical structure from the top layer to the bottom layer. A hierarchical alternating direction multipliers method is presented to solve the overall optimization problem layer by layer in a recursive manner. An improved Nesterov acceleration technique is further introduced to improve the convergence rate. Two cases with different cooling load conditions are conducted on a simulated complex HVAC system to validate the performance of the proposed framework by comparing with three conventional control methods. Results show that it achieves a significant energy saving of 4.05% (275.95 kWh) in case 1 and 5.33% (554.99 kWh) in case 2 compared with the centralized control method. Moreover, it can reduce the computation time by 69.55% averagely in the two cases compared to the conventional distributed control method.