In recent years, maintaining optimal indoor environmental conditions in heritage buildings has received significant attention, yet balancing conservation needs for structures and artworks with occupant thermal comfort often increases energy demand. This study examines whether intelligent HVAC controllers can better balance heritage conservation, occupant thermal comfort and energy use in historic buildings. It focuses on maintaining conservation-oriented temperature and relative humidity bands while limiting HVAC operation time in a UNESCO World Heritage church during the cooling season.
Indoor air temperature, relative humidity, occupancy, and HVAC runtime were monitored at 15-min intervals during the 2023 cooling season at Mission Concepción church, San Antonio (USA). Six data-driven HVAC controllers were developed using artificial neural networks and fuzzy inference systems, configured for joint temperature and relative humidity control and independent temperature-only and humidity-only control. Inputs included indoor temperature, relative humidity and occupancy; output was HVAC runtime. A support vector regression model (80–20 training–validation split), validated using MAE and MAPE, replicated actual building microclimate and enabled offline testing. Performance was benchmarked against a conventional on/off thermostat under identical conditions.
Results show that the intelligent controllers, particularly the FIS-based models, reduce the frequency and duration of deviations from conservation-oriented temperature and relative humidity bands while improving thermal comfort. Compared with the conventional on/off controller, they also reduce HVAC operating time by up to 30% during the monitored cooling season, indicating potential for simultaneous energy savings and enhanced preservation performance.
The study operationalizes “smart preservation” by explicitly treating conservation requirements and thermal comfort as concurrent HVAC control objectives in a real heritage church. It integrates occupancy as a control variable and employs a data-driven SVR microclimate surrogate to evaluate multiple AI-based controllers under identical boundary conditions. This work demonstrates how classical, interpretable ANN and FIS approaches can be leveraged to deliver energy-efficient, conservation-compliant control tailored to historic buildings, bridging a gap between preservation practice and advanced building control strategies.
