Smart Thermostat Development and Validation on an Environmental Chamber Using Surrogate Modelling

The significant contribution of buildings to the global primary energy consumption necessitates the application of energy management methodologies at a building scale. Although dynamic simulation tools and decision-making algorithms are core components of energy management methodologies, they are often accompanied by excessive computational cost. As future controlling structures tend to become autonomized in building heating layouts, encouraging distributed heating services, the research scope calls for creating lightweight building energy system modeling as well monitoring and controlling methods. Following this notion, the proposed methodology turns a programmable controller into a smart thermostat that utilizes surrogate modeling formed by the ALAMO approach and is applied in a 4-m-by-4-m-by-2.85-m environmental chamber setup heated by a heat pump. The results indicate that the smart thermostat trained on the indoor environmental conditions of the chamber for a one-week period attained a predictive RMSE of 0.082–0.116 °C. Consequently, it preplans the heating hours and applies preheating controlling strategies in real time effectively, using only the computational power of a conventional controller, essentially managing to attain at least 97% thermal comfort on the test days. Finally, the methodology has the potential to meet the requirements of future building energy systems featured in urban-scale RES-based district heating networks.