UC Berkeley

There is a growing demand for more energy efficient buildings. Integrated systems with more intelligent controls are an important part of meeting this demand. Model predictive control (MPC) is an established control technique in other fields and holds promise for improved supervisory control in buildings. It has been receiving increasing attention in buildings research but has yet to find its way into common practice. This is due, at least in part, to a mismatch between the tools and techniques used in most MPC development and the tools, skills and processes commonly found in building design and operation. This dissertation investigates an approach to optimization-based control that uses common building simulation tools and could fit more readily into building design and operation practices. Instead of solving optimization problems in real-time to determine control set-points given current states and predicted disturbances, the optimal set-points are pre-computed offline over a grid of possible conditions and the resulting lookup table is used with linear interpolation for control. The feasibility and range of applicability of this approach are evaluated, including analyses of the performance impacts of grid spacing and techniques for problem dimensionality reduction. Three abstract case studies and two detailed case studies are presented. The approach is found to be feasible for supervisory control problems that can be effectively simplified to functions of roughly 5-6 conditions variables, and the case studies show good performance relative to online MPC. The benefits for ease of implementation are significant, but the most useful aspect is likely the feedback it can provide to the design process.