Energy-Efficient Thermal Comfort Control in Smart Buildings via Deep Reinforcement Learning

  • 2019-01-15 07:36:55
  • Guanyu Gao, Jie Li, Yonggang Wen
  • 2

Abstract

Heating, Ventilation, and Air Conditioning (HVAC) is extremelyenergy-consuming, accounting for 40% of total building energy consumption.Therefore, it is crucial to design some energy-efficient building thermalcontrol policies which can reduce the energy consumption of HVAC whilemaintaining the comfort of the occupants. However, implementing such a policyis challenging, because it involves various influencing factors in a buildingenvironment, which are usually hard to model and may be different from case tocase. To address this challenge, we propose a deep reinforcement learning basedframework for energy optimization and thermal comfort control in smartbuildings. We formulate the building thermal control as a cost-minimizationproblem which jointly considers the energy consumption of HVAC and the thermalcomfort of the occupants. To solve the problem, we first adopt a deep neuralnetwork based approach for predicting the occupants' thermal comfort, and thenadopt Deep Deterministic Policy Gradients (DDPG) for learning the thermalcontrol policy. To evaluate the performance, we implement a building thermalcontrol simulation system and evaluate the performance under various settings.The experiment results show that our method can improve the thermal comfortprediction accuracy, and reduce the energy consumption of HVAC while improvingthe occupants' thermal comfort.

 

Quick Read (beta)

loading the full paper ...