Scholarlink

<p>Respiratory tract infections (RTIs) pose a significant threat to human health. Inpatient wards face greater risks of RTIs than restricted or public areas in healthcare facilities. While mechanical or hybrid ventilation systems meet the stringent infection prevention and control (IPC) standards required for inpatient wards, natural ventilation remains the preferred option in low- and middle-income countries (LMICs) due to resource limitations. This paper employed reinforcement learning (RL) to actively control the opening and closing behavior of sliding windows in inpatient wards using a single-sided, single-opening natural ventilation strategy. Simulations were conducted in Guangzhou and Kunming, two Chinese cities with climates suitable for natural ventilation, to compare four window operation strategies: RL-based operation, always open windows, always closed windows, and random operation. Monthly simulations and annual simulations revealed that the RL-based window operation strategy effectively regulates comfort and IPC metrics. This approach enhances natural ventilation’s potential, reducing technical complexity, capital costs, and energy consumption. It is thus an ideal solution for hospital wards in LMICs during design or renovation phases.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reinforcement learning-driven natural ventilation optimization: Reducing respiratory infection risks through adaptive window control in hospital wards of LMICs

  • 期刊论文

  • 发表日期：

    2025年12月23日

查看全文
  • Hao Xie,
  • Pei Zhang,
  • Sui Li,
  • Yang Li,
  • Qi Zhang

摘要

Respiratory tract infections (RTIs) pose a significant threat to human health. Inpatient wards face greater risks of RTIs than restricted or public areas in healthcare facilities. While mechanical or hybrid ventilation systems meet the stringent infection prevention and control (IPC) standards required for inpatient wards, natural ventilation remains the preferred option in low- and middle-income countries (LMICs) due to resource limitations. This paper employed reinforcement learning (RL) to actively control the opening and closing behavior of sliding windows in inpatient wards using a single-sided, single-opening natural ventilation strategy. Simulations were conducted in Guangzhou and Kunming, two Chinese cities with climates suitable for natural ventilation, to compare four window operation strategies: RL-based operation, always open windows, always closed windows, and random operation. Monthly simulations and annual simulations revealed that the RL-based window operation strategy effectively regulates comfort and IPC metrics. This approach enhances natural ventilation’s potential, reducing technical complexity, capital costs, and energy consumption. It is thus an ideal solution for hospital wards in LMICs during design or renovation phases.