An integrated CFD and CNN-GRU approach for accurate prediction of PM2.5 in greenhouse environments of Northeast China

作者：Shuo Zhang, Li Guo, Haiye Yu, Bo Zhang, Meichen Chen, Lei Zhang, Dawei Li, Yuanyuan Sui · 发表于：Building and Environment · 年份：2025 · DOI：10.1016/j.buildenv.2025.113423 · 被引用次数：5 · 研究领域：Air Quality Monitoring and Forecasting、Air Quality and Health Impacts、Urban Heat Island Mitigation

Greenhouses are vital for ensuring food security, improving agricultural productivity, and promoting sustainable crop production through controlled environmental conditions. Accurate and rapid prediction of the spatiotemporal distribution of particulate matter, particularly PM 2.5 (particles with an aerodynamic diameter ≤ 2.5 μm), forms the scientific basis for real-time environmental control systems, thereby improving air quality management and promoting optimal crop growth. In this study, experimental monitoring was combined with computational fluid dynamics (CFD) simulations to analyze the spatiotemporal distribution of PM 2.5 in both solar and glass greenhouses under typical actual production scenarios in Northeast China. Additionally, a machine learning model tailored for PM 2.5 prediction in greenhouse environments was developed. Results revealed that the PM 2.5 concentrations within greenhouses exhibited distinct seasonal characteristics, primarily influenced by variations in ventilation strategies across seasons. Indoor PM 2.5 concentrations were strongly correlated with outdoor PM 2.5 levels and indoor wind speed. CFD simulations revealed that the coupling effect between airflow and greenhouse structure, along with the influence of crop canopies on PM 2.5 suspension and deposition, significantly impacted the spatiotemporal distribution of PM 2.5 within the greenhouses. The Convolutional Neural Network - Gated Recurrent Unit (CNN-GRU) model demonstrated robust predict...