Deep Learning-Based Indoor Air Quality Forecasting Framework for Indoor Subway Station Platforms.

Area of Science:

• Environmental Science
• Public Health
• Computer Science

Background:

• Particulate matter (PM) exposure in enclosed public spaces like subway stations poses significant health risks.
• Continuous monitoring and control of indoor air quality (IAQ) are crucial for protecting commuters and staff.
• Predictive IAQ models are needed to optimize subway ventilation systems.

Purpose of the Study:

• To develop an effective hybrid deep learning framework for forecasting PM10 and PM2.5 concentrations on subway platforms.
• To improve the accuracy of indoor air quality predictions in subway environments.
• To provide a basis for predictive control of subway ventilation systems.

Main Methods:

• Developed a hybrid deep learning framework integrating Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Deep Neural Network (DNN) models.
• Utilized historical indoor and outdoor air quality data for training the hybrid CNN-LSTM-DNN model.
• Evaluated the model's forecasting performance by comparing it against existing deep learning approaches.

Main Results:

• The proposed hybrid CNN-LSTM-DNN model demonstrated effectiveness in forecasting future PM10 and PM2.5 levels.
• The hybrid framework successfully captured temporal patterns and informative characteristics from air quality data.
• The model showed superior performance compared to standalone deep learning models in PM forecasting.

Conclusions:

• The hybrid CNN-LSTM-DNN framework offers a credible and effective solution for predicting particulate matter in subway environments.
• Accurate PM forecasting can enable proactive control of subway ventilation systems, enhancing commuter health.
• This approach advances the application of deep learning in managing indoor air quality in public transportation hubs.