Real-time Personalized Comfort Control Algorithm utilizing User Location and Machine Learning-based PMV Prediction

This study proposes an automated comfort control algorithm that predicts the Predicted Mean Vote (PMV) in real-time and optimizes thermal comfort using only existing air conditioner operation data and radar-based user location information, without requiring additional environmental sensors or wearable devices. Based on data from 66 experimental cases collected in a residential environmental chamber, a linear regression model was developed to correlate air conditioner operating states with the local thermal environment of the occupant. The model predicted local temperature and air velocity with Root Mean Square Errors (RMSE) of 0.72°C and 0.067 m/s, respectively. The developed prediction model is utilized to estimate the occupant's thermal load in real-time, facilitating a multi-phase control logic consisting of 'Rapid cooling' and 'Comfort cooling' modes Environmental chamber experiments confirmed that the proposed algorithm significantly reduces the time required to reach the comfort zone (PMV ±0.5) and enhances occupant thermal satisfaction compared to conventional control methods. T his approach demonstrates the feasibility of user-location-based personalized comfort control for residential air-conditioning systems.