A distributed anytime algorithm for maximizing occupant comfort

We propose a distributed, anytime optimization algorithm to maximize the thermal comfort of building occupants. We consider the building as a set of areas consisting of zones, which are coupled by the capacity of the HVAC equipment as well as the energy and mass balance relations that govern the building dynamics. The resulting non-convex, large-dimensional, constrained optimization formulation is decomposed into area-level subproblems that are solved by distributed agents. At each timestep, the agents cooperate to converge to an equilibrium solution that determines the optimal values of the building operational variables, such as temperature and rate of air flow, that maximizes the total comfort. Our experimental results show that the distributed algorithm (i) is more scalable than the centralized optimization algorithm; (ii) produces a locally optimal solution that is comparable to that resulting from the centralized approach; and (iii) yields a feasible solution even if pre-empted before equilibrium is attained.