Model Predictive Control, Distributed Control, Multi-agent Control, Multi Zone Coordination
Model Predictive Control (MPC) based approaches have recently seen a significant increase in applications to the supervisory control of building heating, ventilation and air-conditioning (HVAC) systems, thanks to their ability to incorporate weather, occupancy, and utility price information in the optimization of heating/cooling strategy while satisfying the physical constraints of HVAC equipment. Many of the proposed MPC solution approaches are centralized ones that often have difficulty in dealing with large-scale building clusters or even a single building consisting of multiple thermal zones due to the high computational cost caused by the large number of decision variables and the overhead in information gathering and distribution. This paper investigates a decomposition technique based on a variant of the Alternating Direction Method of Multipliers (ADMM) that can significantly reduce the computational and communication costs of centralized solutions, and more importantly, facilitates a plug-and-play implementation. The proposed Distributed Model Predictive Control (DMPC) framework takes advantage of parallel computation and collaborations among multiple agents, each of which is assigned to address a smaller dimensional optimization problem. The proposed method is general in that it can accommodate some fairly general types of couplings in agent dynamics as well as in the cost function; therefore it can be used for a broad class of HVAC optimal operation problems. A case study on one of the Purdue Living Labs is carried out to demonstrate the effectiveness of the proposed method. The Living Lab considered is an open office space served by one central air handling unit (AHU) with multiple diffusers whose openings can be controlled individually. The office space is partitioned into multiple thermal zones with individual thermostat controls. In view of significant thermal couplings due to direct air exchange and noticeable load gradient between zones, a multiple thermal zone coordination problem is formulated with the objective of optimally scheduling the different thermostat setpoints for energy minimization and comfort delivery while satisfying actuation constraints. Preliminary results show that the proposed method successfully converges to the optimal solution for the problem concerned in this case study. The optimal solutions demonstrate significant opportunities of inter-zonal coordination and energy savings potentials.