(687b) Modeling and Model Predictive Control Strategies for Building Energy Management

Cara Rose Touretzky and Michael Baldea

Department of Chemical Engineering

The University of Texas at Austin, 1 University Station C0400, Austin, TX 78712

Email:  CaraTour@utexas.edu, MBaldea@che.utexas.edu

Residential and commercial buildings account for over 40% of primary energy consumption in the U.S., with heating and cooling systems accounting for the majority of this figure [1].  While novel materials, new architectural designs, and the availability of alternative energy sources have contributed to reducing building electricity consumption, there still remains a significant amount of unrealized energy and cost savings that can be captured through the optimal control and management of day-to-day building energy use. Making efficient use of energy resources in the presence of fluctuations in weather, occupancy and cost involves complex decisions using both continuous and discrete decision variables, the latter corresponding to the on/off state of various pieces of equipment, e.g., heaters and chillers. The inherent complexity of the building energy management process is thus a natural candidate for optimization [2, 3].

The practical implementation of an optimization-based control and energy management approach is, however, hindered by the considerable size and complexity of the mathematical models representing the dynamics of a building; such models do not lend themselves easily to use in online calculations. An additional complication originates in the building dynamics: the significant discrepancy in the thermal mass of the air circulated in by the Heating Ventilation and Air Conditioning (HVAC) and the thermal mass of the building itself, along with the presence of energy recovery devices (e.g., air preheating using the exhaust air), give rise to a dynamic behavior that evolves over multiple time scales [2]. In this context, the evolution of local (e.g., room-level) temperatures is fast, while the total energy stored in the building evolves over a much longer time horizon. In turn, this multiple time scale behavior is at the origin of the stiffness of the mathematical models of buildings. Using such for controller design is impractical: the resulting controllers are almost invariably difficult to tune and sensitive to modeling errors and noise.

Motivated by the above, we propose a novel model reduction and energy management framework for buildings. We demonstrate that, owing to energy recovery features such as air preheating, air recirculation and thermal energy storage, the dynamic behavior of a building is similar to that of an integrated chemical process. Based on this analogy, we expand on our recent results [4] to develop a method for deriving a reduced-order, nonlinear model that describes the dominant, slow dynamics of a building, reflecting the influence of the large thermal mass of the system. Subsequently, we introduce a set of techniques that enhance model portability, i.e., facilitate a fast, optimal identification of the parameters of the model from building operation data, and the deployment of the model in specific building applications. The resulting models are low-dimensional and non-stiff, and thus well suited for control applications. Using the nonlinear model predictive control (NMPC) paradigm, we develop a new approach to building energy management based on the aforementioned reduced-order models. The stability properties of the proposed NMPC scheme are proved theoretically. Finally, we provide a set of experimental results obtained at the University of Texas at Austin Thermal Façade Laboratory, demonstrating the effectiveness of the proposed building energy management scheme.

References:

[1] Buildings Energy Data Book Table 1.1.3, U.S. Department of Energy (2011)

[2] Ma, Y., Anderson, G., Borrelli, F. A Distributed Predictive Control Approach to Building Temperature Regulation.  American Control Conference (2011)

[3] Wang, S., Ma, Z.  Supervisory and Optimal Control of Building HVAC Systems: A Review.  HVAC&R Research (2008)

[4] Baldea, M., Dautidis, P. Dynamics and Nonlinear Control of Integrated Process Systems, Cambridge University Press (2012)