As of 2011 buildings consumed 41% of all primary energy in the U.S. and can represent more than 70% of peak demand on the electrical grid. Usage by this sector has grown almost 50% since the 1980s and projections foresee an additional growth of 17% by 2035 due to increases in population, new home construction, and commercial development. Three-quarters of building energy is derived from fossil fuels making it a large contributor of the country's CO2 and NOx output both of which greatly affect the environment and local air quality. Up to half of energy used by the building sector is related to Heating, Ventilation, and Air-Condition systems. Focusing on improving building HVAC control therefore has a large aggregate effect on US energy usage with economic and environmental benefits for end users. This dissertation develops cascaded loop architectures as a solution to common HVAC control issues. These systems display strong load-dependent nonlinearities and coupling behaviors that can lead to actuator hunting (sustained input oscillations) from standard PI controllers that waste energy and cost money. Cascaded loops offer a simple way to eliminate hunting and decouple complex HVAC systems with minimal a priori knowledge of system dynamics. As cascaded loops are easily implementable in building automation systems they can be readily and widely adopted in the field. An examination of the current state of PI control in HVAC and discussion of coordinated, optimal control strategies being developed for reduced energy usage are discussed in Chapter 1. The following two chapters outline the structure and benefits of the cascaded architecture and demonstrate the same using a series of simulation case studies. Implementation approaches and parameterizations of the architecture are explored in Chapter 4 with a derivation showing that the addition of an additional feedback path (i.e., inner loop control) provides more design freedom and ultimately allows for improved control. Finally, Chapter 5 details results from initial cascaded loop implementation at three campus buildings. Results showed improved control performance and an elimination of identified hunting behavior.
