Enhanced Air Quality Prediction Using a Coupled DVMD Informer-CNN-LSTM Model Optimized with Dung Beetle Algorithm.

Area of Science:

• Environmental Science and Engineering
• Atmospheric Science
• Artificial Intelligence in Environmental Monitoring

Background:

• Accurate air quality index (AQI) prediction is vital for environmental assessment due to the complex, nonlinear nature of air quality data.
• Traditional models face limitations in feature utilization, parameter tuning, and prediction accuracy, necessitating advanced approaches.

Purpose of the Study:

• To develop an enhanced AQI prediction model by optimizing Variational Mode Decomposition (VMD) parameters using the Dung Beetle Algorithm (DBO).
• To integrate the Informer and Convolutional Neural Network-Long Short Term Memory (CNN-LSTM) models for adaptive sequential prediction.
• To improve the accuracy and efficiency of AQI forecasting.

Main Methods:

• Feature selection using the correlation coefficient method from meteorological data.
• DBO-based optimization of VMD parameters (penalty factor, number of modes).
• Hybrid prediction using Informer for high-frequency and CNN-LSTM for low-frequency data components, followed by reconstruction.

Main Results:

• The proposed DBO-VMD-Informer-CNN-LSTM model significantly outperforms existing methods in Beijing's air quality data.
• Demonstrated improvements include a 13.59% decrease in Mean Absolute Error (MAE), a 7.04% decrease in Root-Mean-Square Error (RMSE), and a 1.39% increase in R-square (R²).
• The DBO algorithm shows higher computational efficiency and accuracy compared to other swarm intelligence algorithms.

Conclusions:

• The developed coupling model offers a superior and more accurate approach to AQI prediction.
• This method addresses limitations of traditional models and provides a novel solution for environmental monitoring.
• The enhanced prediction accuracy contributes to better atmospheric environment assessment and management.