Machine Learning Algorithm for Optimising Comfort Cooling in Buildings

Authors- Shashank R, Professor Dr. M. Rajagopal

Abstract-Optimizing comfort cooling in commercial buildings is essential for reducing energy consumption while ensuring occupant comfort. Traditional HVAC systems often operate on fixed schedules and static parameters, leading to inefficiencies. This paper presents a machine learning (ML)-based approach to dynamically optimize HVAC cooling operations. The proposed system integrates predictive algorithms for cooling load forecasting and reinforcement learning (RL) for real-time HVAC control optimization. Real-time sensor data (temperature, humidity, occupancy) and external inputs (weather forecasts, time-of-day) are used to predict cooling demand. A regression model forecasts cooling loads, while RL algorithms optimize HVAC actions, such as adjusting compressor and fan speeds and air distribution and, to minimize energy use while maintaining thermal comfort. The RL model’s reward function penalizes energy overuse and deviations from comfort thresholds (e.g., 22-25°C temperature, 40-60% humidity). The results indicate that the ML-based system significantly lowers energy consumption compared to traditional control methods, without sacrificing comfort. The model adapts to real-time changes in building usage and external conditions, offering a scalable, flexible solution for smart building management. This method improves energy efficiency while also prolonging the lifespan of HVAC components by reducing unnecessary wear. The proposed method presents a promising framework for integrating advanced ML techniques, such as predictive analytics and real-time optimization, into building management systems, contributing to sustainability through reduced carbon emissions and dynamic adaptation to occupant preferences.

DOI: /10.61463/ijset.vol.13.issue1.150