Predictive Control Method for Comfort and Thermal Energy Enhancement in Buildings

HVAC systems represent the most significant energy consumers in buildings, constituting over 60% of total energy usage. This research endeavors to enhance energy efficiency and thermal comfort within buildings, particularly focusing on heating and cooling systems. To optimize energy performance, we have developed a model predictive control system leveraging artificial neural networks. Our implementation incorporates cutting-edge technologies such as field-programmable gate arrays (FPGAs), communication modules and protocols (MQTT), for enhanced scalability and adaptability. Our study highlights a significant reduction in heating and cooling sensible thermal energy consumption with the adoption of Neural Network Model Predictive Control on the PYNQ board. It resulted in an exceptional decrease of 40.8% and 37.8% in the yearly energy requirements for heating and cooling, respectively, when compared to conventional On/Off control systems.