基於深度強化學習之室內空氣品質控制系統研究

Abstract

近年來世界各地飽受新冠肺炎的侵擾，許多政府為了防止病毒的擴散頒布了一系列措施以降低傳染病毒的風險，但是有些措施在一些場合無法完全的實施，如於學校中保持安全距離、在餐廳中配戴口罩等等，因此本研究嘗試以不干預人們行為的方式降低感染病毒的風險以及保持空間內的舒適和省電的條件下建立一套自動控制室內空氣品質的系統。該系統稱之為室內空氣品質自動控制系統(Indoor Air Quality Auto Control, IAQAC)，藉由深度強化學習(Deep Reinforcement Learning)技術使系統能夠自行找出對該空間最佳的策略以保持舒適、低感染風險、省電的效果；另外研究中使用了遷移學習先後在模擬環境和現實環境中訓練以降低過多的時間成本；最後搭建了一套自動收集資料、控制設備的系統以提供必要的資訊和執行系統決策的動作。

In recent years, the world has been affected by COVID-19. Many governments have enacted a series of measures to prevent the spread of the virus to reduce the risk of infection, but some measures cannot be fully implemented in some situations, such as keeping a safe distance in schools, wearing masks in restaurants, etc. Therefore, this study attempts to create a system that automatically controls indoor air quality without interfering with people's behavior to reduce the risk of virus infection and to keep the space comfortable and energy efficient. The system, called Indoor Air Quality Auto Control (IAQAC), uses Deep Reinforcement Learning (DRL) techniques to enable the system to find its own optimal strategy for the space to maintain comfort, low risk of infection, and energy savings. In addition, Transfer Learning was used to train the system in both simulated and realistic environments to reduce excessive time costs; finally, a system was built to automatically collect data and control equipment to provide necessary information and perform system decision making actions.