Clarence Agbi Defense

Clarence Agbi Defense

Starts at: May 5, 2014 9:00 AM

Ends at: 12:00 PM

Location: PH B34


In the wake of rising energy costs, there is a critical need for sustainable energy management of commercial and residential buildings. Buildings consume approximately 40% of total energy consumed in the US, and current methods to reduce this level of consumption include energy monitoring, smart sensing, and advanced integrated building control. However, the building industry has been slow to replace current PID and rule-based control strategies with more advanced strategies such as model-based building control. This is largely due to the additional costs of modeling, the need for regular maintenance, and general uncertainty that these controllers are robust enough to be used in real conditions. This dissertation seeks to address each of these issues, and proposes novel strategies to improve the identification of building models used for control and strategies to make model-based building control more robust to model uncertainties.

The first half of this thesis discusses the issue of constructing and identifying grey-box building models for control. Although grey-box building models are flexible given their parametric structure, the parameters in the model are also harder to identify because of the complexity of the structure. Furthermore, current approaches towards building model identification are not scalable for large building models. Therefore, we use notions of model identifiability to characterize the quality of a building model structure, and to determine the parameters in the building model that may be unidentifiable. Based on this information, we can strategically redesign the building model structure to improve the identifiability of the model. We also present a decentralized identification scheme to reduce the computational effort and time needed to identify large building models.

The second half of this thesis discusses the challenge of using building models to control a building environment. Specifically, we address the issue of building models that do not exactly match the dynamics of the building, and the impact of model uncertainty on the performance of a model-based building controller such as model- predictive control. We propose a robust control strategy that builds on H∞ control, and we introduce a tuning law to increase the accuracy of the controller in the presence of high model uncertainty. Finally, we propose and demonstrate a hierarchical control framework for robust model-based control in building environments.