Building Energy Simulation Using Constrained Optimization

Abstract

This thesis presents an equation-based approach to steady state building energy modeling using constrained optimization. Models using this approach can be built from algebraic equations that describe a building and its HVAC equipment and knowledge of the system’s overall control strategy. The equation-based nature of this approach allows arbitrary systems to be represented, and simple rules can be followed to construct models that produce a valid energy balance overall operating conditions, including atypical scenarios such as temperatures drifting above or below their desired set points. The theory behind constrained optimization modeling of steady state HVAC systems is presented, and methods for constructing typical HVAC systems are given. Mathematical methods for integrating dynamic effects such as thermal mass into these models are also presented. To efficiently solve these models, a prototype program named Beryl is presented along with a performance analysis of the computer resources required for different types of modeling. Finally, future areas of research are given such as modeling improvements and improvements to numerical solvers for the specialized problems being solved. The solver developed in this work was able to perform yearly building energy simulations on single AHU in 0.5 to 5.0 seconds using hourly time steps on a single thread. Factors that affect simulation runtime were also analyzed. Based on these results and the robustness of the simulations, constrained optimization based building energy modeling was shown to be a valid and potentially useful approach.