Development of optimal thermostatic control algorithms for efficient energy use and peak load reduction in buildings.
For a large building, number of thermal zones rapidly increases, and thus the resulting large-scale centralized control problem becomes infeasible to implement in low-cost embedded platforms used in most of the HVAC control equipment. In this project, we develop a distributed model predictive controller (MPC) that (i) minimizes peak cooling load on the HVAC equipment (ii) while achieving a satisfactory thermal comfort inside the building, and (iii) is capable of handling large-scale systems with uncertain dynamics (e.g. ambient air-temperature, internal heat gains, etc.). The key difference of our approach compared to previous approaches that utilize MPC is that in our approach the Air Handling Units (AHU) are not controlled directly. Instead, we control the AHUs are indirectly by sending control inputs to the zone thermostats, and require minimal hardware changes. The thermal dynamics of an individual zone will be modeled by a lumped resistance-capacitance circuit. The model parameters are physics-based, and exact calculation of these parameters requires comprehensive details of the zones. Thus, we estimate these parameters from sensor measurements using system identification methods.