| Commercial heating, ventilation, and air conditioning (HVAC) systems have been the focus of sustainability research due to their large energy footprint and relatively rudimentary control strategies. This thesis presents a novel approach to building thermal control which utilizes weather forecasting and historic data to apply an iterative learning control (ILC) solution. ILC is shown to reject repeating disturbances and produce tight target trajectory tracking, so that setpoints may be strategically chosen to conserve energy and preemptively account for future disturbance. Furthermore, our approach does not require detailed model information, in contrast to, for example, the popular model predictive control (MPC) method.;For simulation and system analysis, we model a multi-zone building as a lumped-parameter thermal resistance-capacitance network. This model is used to perform simulations and to study the inherent passivity in multi-zone building systems. We show the system to be strictly output incremental passive (SOIP) under closed loop negative feedback, which in turn implies convergence over iterations of ILC. For the ILC application, we match predicted disturbances to a historical database, and use well-matched days for learning. We also present a modified version of the algorithm to control both temperature and humidity using a single input, the mass flowrate. This algorithm is shown to converge to a family of solutions lying on the boundary of a set, the location and geometry of which can be designed to maintain comfort and reduce energy consumption.;For experimental valuation, we have designed, built, and instrumented a unique test facility. This intelligent building testbed was created to serve as a standardized platform for building modeling and control evaluation. Key features include wireless control and sensing of temperature, humidity, energy, and solar effects; and a modular structure with multiple zones and adjustable thermal and mass coupling. A key feature of this testbed is its placement inside a thermally-controlled enclosure, which can generate repeatable conditions. Simulation and experimental evaluations of the proposed modeling and control methods show improved setpoint tracking. |
