Domestic heating systems exhibit significant energy flexibility when integrated with heat pumps and hot-water buffer tanks, especially with fluctuating day-ahead energy prices. However, optimizing these systems in large buildings with shared resources poses crucial challenges. While most existing methods target single-room or single-family houses, this study delves into complexities within a three-story building housing six apartments. The building's heating system comprises a hot water buffer tank, mixing loop, floor heaters, and a Ground Source Heat Pump (GSHP) controlled by a weather-compensated control strategy (WCS). Our approach aims to tackle challenges like integrating real sensor data, scalability, varying weather effects, and diverse resident heat use preferences. We employ CTSMR software to identify thermal behaviour and use reinforcement learning to design an intelligent Uppaal Stratego controller. Our results reveal a 43% reduction in energy costs while maintaining comfort levels compared to WCS. The temporal validity of estimated thermal models is also analyzed.
Read full abstract