PM_2.5 concentration prediction using a whale optimization algorithm based hybrid deep learning model in Beijing,...

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Background:

• Particulate Matter (PM2.5) is a major global air pollutant affecting visibility, climate, and public health.
• Accurate PM2.5 concentration prediction is crucial for risk assessment and early warning systems.

Purpose of the Study:

• To develop and evaluate a novel hybrid machine learning model for predicting daily PM2.5 concentrations.
• To enhance the accuracy and reliability of short-term and medium-term PM2.5 forecasting.

Main Methods:

• A hybrid model combining Whale Optimization Algorithm (WOA), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Attention Mechanism (AM) was developed.
• The model was trained and tested using daily meteorological and air pollution data from 2014 to 2018.
• SHAP analysis was employed to identify key predictors of PM2.5 concentrations.

Main Results:

• The WOA-CNN-LSTM-AM model significantly reduced prediction errors compared to standalone CNN and LSTM models.
• Achieved MAE of 14.29, RMSE of 21.96, MBE of -0.23, and R² of 0.93.
• Demonstrated 30%-54% higher accuracy in medium-term forecasts compared to WOA-CNN-LSTM and 26%-39% over CNN-LSTM-AM.

Conclusions:

• The proposed hybrid model offers a reliable and accurate tool for daily PM2.5 prediction.
• NO2 and CO were identified as primary drivers influencing PM2.5 concentrations.
• The model supports effective air pollution control strategies through improved forecasting.