Enhancing thermal comfort and sustainability in building energy management through edge machine learning-enabled smart controllers

Abstract

The concept of smart cities is becoming increasingly popular, driven by the goal of improving residents’ overall quality of life. Effective thermal management is a crucial component of smart cities, especially in light of significant changes in global weather patterns. These changes pose challenges for individuals seeking indoor comfort, particularly in workplaces and schools. Occupants often struggle with extreme temperatures, leading to increased energy consumption due to excessive use of air conditioners and heaters. In this work, we propose a novel edge machine learning-based smart controller for air conditioners to maintain user-preferred thermal comfort levels while minimizing energy consumption. The trained long short-term memory model is embedded in a mobile application designed to predict the predicted mean vote (PMV) based on temperature setpoints and the PMV index. Through this mobile application, personalized parameters such as clothing insulation, metabolic rate, and other environmental factors are communicated to the controller. The performance of the smart controller in maintaining PMV-based temperature setpoints for cooling is compared with fixed setpoints. The results reveal that by utilizing PMV-based setpoints for male occupants, the smart controller reduces energy consumption by 38.14% compared to a fixed setpoint of 26°C and by 60.68% compared to 25°C, while maintaining the user-preferred thermal comfort level.