Data-driven assessment of humidity-adaptive thermostat control for energy efficiency and thermal comfort in arid climates

Space cooling in arid climates is characterized by a disproportionately high latent load. A psychrometric decomposition of indoor air enthalpy across five experimental datasets shows that humidity-driven latent enthalpy accounts for 43–48% of total cooling energy, even though conventional thermostats regulate only dry-bulb temperature. Conventional control therefore leaves a near-equal share of the cooling load unaddressed. To address this gap, this study introduces the Heat Index Thermostat (HiT), a setpoint-level control strategy that adjusts the commanded dry-bulb temperature as a linear function of indoor relative humidity, without modifying dehumidification equipment or refrigerant circuits. HiT is designed as a software-only retrofit that can be implemented on existing smart thermostats through the standard setpoint interface, requiring no additional sensors, cloud connectivity, or HVAC hardware changes. Further, system performance is evaluated using the Comfort-Energy Efficiency Ratio (CEER), which captures comfort satisfaction per unit of electrical demand. Random forest feature attribution (using SHAP), applied retrospectively as an offline evaluation tool rather than as a component of the real-time control logic, verified that indoor Relative Humidity was the most significant factor influencing HVAC power consumption (mean |SHAP| = 0.501). The field experiment conducted at a 23°C base temperature demonstrated an improvement of 90.6% in CEER with respect to traditional controls (p < 0.001, Cohen’s d = 0.79), a reduction in mean PPD from 19.75% to 15.47%, and a 25.2% decrease in mean power consumption. These results establish HiT as a practical, deployable approach for demand-side management DSM in arid-climate buildings.