Intelligent Control of a Villa AC Unit

The primary objective of this study was to assess potential energy savings achieved by incorporating the thermal comfort concept, calculated through the predictive mean vote (PMV) index as outlined in ASHRAE Standard 55-2020, to control the operation of an air conditioning (AC) unit in a Kuwaiti villa, as opposed to an AC with conventional thermostat based on room air temperature (TA). The study introduced an innovative technique by leveraging artificial intelligence (AI) and machine learning (ML) algorithms for the estimation of the thermal comfort, specifically the mean radiant temperatures (MRT), as opposed to gauging the same temperature by the wet-bulb globe temperature (WBGT) meter.
Upon examination of its mathematical equation, thermal comfort becomes apparent that MRT, coupled with TA, stands out as the most significant variables in these calculations. Moreover, MRT is the only parameter that requires complicated sensors to measure. In comparison, the AI and ML algorithms only take four parameters to properly predict the MRT: outside temperature, inside temperature, thermostat set temperature, and time. The algorithms presented a viable alternative, enabling highly accurate predictions of PMV values.
Among three ML models examined, the artificial neural network demonstrated the lowest unmet ratio of PMV values (7%) and yielded the highest energy savings (26%). These findings affirm the artificial neural network as the optimal ML model among the three for AC system control.