Data Driven Q-Learning for Commercial HVAC Control

Abstract

Commercial HVAC systems play a key role in the consumption of commercial buildings. Therefore, there is a need for a safe and cost-effective HVAC control algorithm. An algorithm that can learn from the previous experience and reduce the associated energy cost is highly required. In this paper, a data driven-based reinforcement learning for the optimal control of HVAC system of a commercial building is proposed. Random forests technique is utilized to provide a data driven model for the HVAC system. A Q-learning algorithm is used as a type of reinforcement learning to minimize the building energy consumption cost while maintaining the comfort level. The results showed that the proposed algorithm is able to maintain the required building temperature and provide a lower energy cost compared to the base case schedule.