Supervisory control for intelligent building systems
Abstract
The objective of an office building is to provide optimum comfort that enables the occupants to attain maximum productivity. To meet this objective, a heating, ventilating, and air-conditioning (HVAC) system provides climate control; however, with an energy cost to the building owner. From an economic point of view it is important to minimize energy consumption while maximizing comfort. A three-level hierarchical control system composed of a supervisor, coordinators, and local controllers was formulated to meet this objective such that the building control system learns the characteristics of the building and determines the best control strategy. This intelligent building system (IBS) contains local controllers that control the equipment, a supervisor that monitors predicted disturbances to anticipate control actions, and coordinators that modify the supervisor's decisions to compensate for unexpected real-time disturbances. The supervisor provides an optimistic plant to operate the building using past information to predict the performance of the building. This level can learn the building dynamics and formulate a model from generic units or "building blocks." These generic units are simple models that can be identified with data from the building and its HVAC system. Since comfort is needed only when the building is occupied, a technique was developed to predict occupancy using motion sensors. This technique requires no prior knowledge and with sufficient data can create a probability of occupancy. To minimize energy cost while maintaining comfort, a cost function was formulated. This functional relates comfort and energy costs into a unified performance index. An experimental test bed, composed of three offices, was used to verify portions of the supervisory control. Motion data was converted into probability of occupancy profiles, weather forecasts were collected, and the building model was identified. Predicted building block coefficients and occupancy profiles showed good correlation with the actual data. This information, with the performance index, was used to determine analytically optimal time dependent set points for a room.
Degree
Ph.D.
Advisors
Shoureshi, Purdue University.
Subject Area
Mechanical engineering|Systems design|Energy
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.