| Current research has shown that building energy consumption contributes to more than 40% of the total energy use in United States. In recent years, the investigation of the reduction of a building's energy consumption, both at the design stage and operational stage, has received great attention. To obtain an optimal design and improve energy efficiency during operation, a reliable building energy model is needed to incorporate the important static and dynamic information from various resources. Static information includes building geometry, materials, and installed HVAC system, while dynamic information includes weather information, occupants' temperature schedule, inputs from sensors (e.g., temperature and occupancy), and internal heat and moisture sources during operation.;At the design stage, using a building energy model can assist architects with normative design and building optimization workflows. The developed energy model in this dissertation can be easily integrated with Revit Architecture, one of the most commonly used building design tools, to accelerate the design optimization process. Its use can save redundant efforts of going back-and-forth between an energy modeling tool and a design tool to assess the energy impact of design modifications. At the operational stage, a machine learning method is utilized to predict the occupants temperature schedule, and this schedule is input to the model-based predictive control for further optimization with the goal of minimizing energy consumption while maintaining comfort conditions.;Though many attempts have been made to study the basic mechanisms of building energy modeling and predictive control, major challenge still lie in how to develop a simple but also relatively accurate thermal model to simulate a building's energy performance. The other challenge lie in how to develop a comprehensive cost function with sufficient constraints for a model predictive control algorithm. This dissertation introduces and illustrates a method for integrated building control to reduce energy consumption and maintain indoor temperature setpoints, based on the prediction of occupant behavior patterns and weather information. It primarily focuses on the following goals. Firstly, An energy model for building energy simulation has been developed, and integrated to assist architects to do design optimizations. Secondly, a Model-Based Predictive Control (MBPC) method has been developed and experiments have been conducted to validate the energy-saving benefits of MBPC compared with other conventional control methods. In the end, a machine-learning algorithm is developed to predict the future temperature schedule based on historic data. |
