Model in-cognizant control of residential HVAC units with limited sensing and actuation.

In this paper, we consider the problem of controlling residential heating, ventilation and air conditioning (HVAC) units in response to changes in grid-side electrical power imbalances causing unacceptable frequency. We derive a novel energy-based model that relates the HVAC physics-based dynamics to both real and reactive power balance at the point of interconnection with the grid. Based on this modeling, we propose a composite control comprising of a robust sliding mode controller in tandem with a slower model predictive controller that can achieve near-optimal physical and economic performance. In contrast to several other approaches in the literature, we analyze whether a limited number of HVAC units can meet the stringent performance metrics set by the ARPA-E/NODES program on following the regulation signal, while maintaining consumer comfort. Theoretical and simulation-based evidence is provided to show that the proposed approach to control a single HVAC unit results in a provable response simultaneously satisfying NODES program performance metrics and consumer comfort constraints. The use of this model overcomes fundamental issues concerning limited sensor measurements and model uncertainties.