Model predictive control (MPC) offers a tremendous scope in optimizing the consumption of energy by building HVAC systems. This paper presents an automated real-time procedure for the development of linear parametric models of building air-conditioning systems through system identification for the implementation of the MPC algorithms. The procedure is used to decide on the various aspects of system identification such as selecting the model structure, the inputs to the system, the interaction of the systems with their neighbors, and the updating of the model coefficients in real-time. The effectiveness of the procedure is demonstrated by modeling the various components air-conditioning systems of a real building. The root mean squared error was used as a performance metric to gauge the models. The paper also demonstrates that a 15 minute sampling interval is sufficient to model the dynamics of the air-handling unit and the room temperatures, but a faster sampling rate may be required to model the VAV boxes.
Survivable Networks via UAV Swarms Guided by Decentralized Real-Time Evolutionary Computation
