SMART GRID COORDINATION IN BUILDING HVAC SYSTEMS

In the context of real time electricity pricing, energy consumption by Heating, Ventilation, and Air Conditioning (HVAC) systems is usually heaviest when prices are at their highest. In order to reduce expenditure while adequately maintaining indoor comfort conditions, Economic Model Predictive Control (EMPC) policies can be implemented in conjunction with Thermal Energy Storage (TES). This equipment con guration allows the time-shift of chiller power consumption away from periods of high demand to periods of low energy cost. Overcoming the main drawbacks intrinsic to EMPC policies constitutes the initial focus of this work. The rst issue concerns the susceptibility of EMPC to disturbance prediction quality, as this algorithm relies on weather and electricity price forecasts to generate control actions. Thus, two gray box models of increasing complexity are developed to illustrate the fundamentals of disturbance forecasting with shaping lters. Additionally, two data driven models are also presented for comparison. The discussion is then expanded to forecasting quality based on the amount of information available, and proper economic evaluation of these scenarios with gray and black box models is performed. The second issue of interest is related to the susceptibility of EMPC to prediction horizon, since this methodology computes control actions based on a receding horizon framework. Usually, control policies that reduce expenditure the most also rely on relatively large horizon sizes. These implementations tend to result in considerable computational burden. While horizon size reductions are commonly used to lessen the associated computational needs, they also carry signi cant economic performance degradations. To solve this issue, a novel Economic Linear Optimal Control (ELOC) capable of enforcing constraints statistically is developed. Additionally, the ELOC feedback is used to generate a receding horizon formulation capable of enforcing point-wise-in-time constraints, termed constrained viii ELOC. This algorithm is virtually insensitive to horizon size while retaining and even surpassing the economic performance of EMPC. The third major achievement of this work is the development of an HVAC equipment sizing and optimization methodology. This goal is achieved through the implementation of a gradient search algorithm designed to predict returns on investment based on a net present value analysis. As part of this technique, the ELOC and constrained ELOC methodologies are expanded for equipment design. The successful implementation and convergence of this numeric optimization is illustrated with a case study.PH.D in Chemical Engineering, December 201