Related Experiment Video
Updated: Jul 8, 2026
08:47
Computer Vision-Based Biomass Estimation for Invasive Plants
Published on: February 9, 2024
Forecasting of PM2.5 concentration based on variational mode decomposition and deep learning
1School of Big Data and Artificial Intelligence, Anhui Institute of Information Technology, Wuhu, 241199, China.
Scientific Reports
|June 4, 2026
Summary
This study introduces an advanced PM2.5 forecasting model using variational mode decomposition (VMD) and deep learning. The novel approach enhances air quality prediction accuracy, crucial for environmental monitoring.
Area of Science:
- Environmental Science
- Data Science
- Artificial Intelligence
Background:
- Accurate PM2.5 forecasting is vital for air quality management.
- PM2.5 time series exhibit complex nonlinearity and non-stationarity.
- Existing models often struggle with these characteristics.
Purpose of the Study:
- To develop a robust forecasting model for PM2.5 concentration.
- To address the challenges of nonlinearity and non-stationarity in PM2.5 data.
- To improve the accuracy and reliability of air quality predictions.
Main Methods:
- Variational Mode Decomposition (VMD) to decompose PM2.5 series into intrinsic mode functions (IMFs).
- Sample Entropy (SE) and K-means clustering to categorize IMFs by frequency.
- A deep learning architecture combining Temporal Convolutional Networks (TCN), Bidirectional Long Short-Term Memory (BiLSTM), and an attention mechanism for forecasting.
Main Results:
- The proposed model achieved superior forecasting accuracy.
- Lowest Root Mean Square Error (RMSE) of 16.920 µg/m³.
- Lowest Mean Absolute Error (MAE) of 11.134 µg/m³ and highest R² of 0.960.
Conclusions:
- The VMD-TCN-BiLSTM model with attention significantly improves PM2.5 forecasting.
- The method effectively handles the complex nature of PM2.5 time series.
- This approach offers enhanced capabilities for air quality monitoring and research.
