Development of a deep reinforcement learning (DRL) based controller that can help to create a healthy and energy-efficient dynamic indoor environment
Energy-efficient building thermal control policies are needed to reduce the operational energy use of HVAC systems in buildings. Indoor temperatures tend to be maintained within narrow limits, which leads to higher operational energy use. Also, long-term exposure to this type of environment might not be optimal for building occupants.
Studies suggest that a dynamic environment may be healthier for the human body. However, there are multiple challenges associated with the implementation of a control policy that factors in important parameters like the energy efficiency of the HVAC System, the dynamic indoor environment, and the thermal acceptability requirements. In addition, it is hard to model the policy because its parameters may differ from case to case.
To tackle this complexity, ICE intends to develop a Deep Reinforcement Learning-based framework (DRL) for energy optimization and healthy thermal environment control in buildings. Since most of the control variables are continuous in buildings, we implement a novel DRL algorithm known as deep deterministic policy gradient (DDPG).