An Experimental Machine Learning Evaluation on Adaptive Thermal Comfort in Hospitals

Abstract

The healthcare sector is a significant contributor to global energy consumption, particularly within the heating, ventilation, and air conditioning (HVAC) market, due to stringent requirements for maintaining indoor thermal comfort for patients, staff, and visitors in hospital wards, rooms, and intensive care units. This study presents a novel approach employing a Machine Learning-based Linear Regression Algorithm to predict indoor adaptive thermal comfort within the inpatient medical wards. The methodology establishes a robust correlation between key indoor environmental parameters including air temperature, relative humidity, and air velocity, and thermal comfort indices, including Predicted Mean Vote (PMV) and Predicted Percentage Dissatisfied (PPD). Real-time measurements of indoor environmental conditions were conducted in selected hospitals in Islamabad, Pakistan, utilizing calibrated sensors to capture ambient temperature, wet bulb globe temperature, relative humidity, air velocity, light intensity, and CO₂ levels. This empirical data was integrated with responses from thermal comfort questionnaires, assessing the perceptions of patients, medical staff, and visitors regarding thermal sensation, acceptability, preference, and overall comfort. The adaptive ML-based, predictive analysis identified optimal thermal comfort ranges for hospital wards, recommending indoor air temperatures between 22.0°C and 23.0°C, relative humidity levels between 50% and 55%, and air velocities between 0.1 and 0.2 m/s. The findings revealed a significant impact of overcooling and undercooling on PMV and PPD levels, emphasizing the need for precise HVAC system control to enhance both energy efficiency and occupant comfort. This research contributes to advancing adaptive thermal comfort modeling in healthcare facilities, offering insights for sustainable HVAC management and improved patient outcomes.