Comfort-aware HVAC Aggregation Method based on Deep Reinforcement Learning

Abstract

Extracting demand-side flexibility is important to utilize renewable energy as much as possible with keeping the grid stable. Heating, ventilation, and air conditioning (HVAC) system takes a necessary role in our comfort life and consumes more than 40% energy consumption of typical office buildings. HVAC systems are known as potential resources to extract demand-side flexibility because of easy performance control. This paper proposes a comfort-aware HVAC aggregation method based on deep reinforcement learning to both maintain a comfortable temperature and extract flexibility. Experiments show the proposed method satisfies the given demand-response requirements and thermal comfort.