USING DEEP LEARNING TO OPTIMIZE HVAC SYSTEMS IN RESIDENTIAL BUILDINGS

Abstract

ABSTRACT HVAC systems are crucial for maintaining indoor temperature and humidity in buildings but consume significant energy, accounting for over 50% of a building’s energy use. This study proposes a deep reinforcement learning (DRL) algorithm for optimizing energy consumption in residential building HVAC systems while maintaining occupant comfort. Climate data was collected using low-cost sensors, and a co-simulation framework was developed for offline training and validation of our DRL-based algorithm. The proposed DRL-based algorithm was compared to a rule-based HVAC system regarding energy consumption and occupant comfort. Results show that the proposed algorithm can reduce energy consumption by up to 15% compared to the rule-based HVAC system. DRL is a suitable approach for optimizing HVAC systems due to its ability to adapt to the dynamics of multi-parameterized systems. This study contributes to sustainable building design by proposing a DRL-based algorithm to reduce energy consumption while maintaining a comfortable indoor temperature. Using low-cost sensors and a co-simulation framework provides a practical and cost-effective method for training and validating the proposed algorithm.