A real-time model predictive control for building heating and cooling systems based on the occupancy behavior pattern detection and local weather forecasting

Current research studies show that building heating, cooling and ventilation energy consumption account for nearly 40% of the total building energy use in the U.S. The potential for saving energy through building control systems varies from 5% to 20% based on recent market surveys. This papers introduces and illustrates a methodology for integrated building heating and cooling control to reduce energy consumption and maintain indoor temperature set-point, based on the prediction of occupant behavior patterns and local weather conditions. Advanced machine learning methods including Adaptive Gaussian Process, Hidden Markov Model, Episode Discovery and Semi-Markov Model are modified and implemented into this study. A Nonlinear Model Predictive Control (NMPC) is designed and implemented in real-time based on dynamic programming. The experiment test-bed is setup in the Solar House, with over 100 sensor points measuring indoor environmental parameters, power consumption and ambient conditions. The experiments are carried out for two continuous months in the heating season and for a week in the cooling season. The results show that there is a 30.1% measured energy reduction in the heating season compared with the conventional scheduled temperature set-points, and 17.8% energy reduction in the cooling season.