Related Experiment Videos
A Novel Method for Regional NO2 Concentration Prediction Using Discrete Wavelet Transform and an LSTM Network
Bingchun Liu1, Lei Zhang1, Qingshan Wang2
1School of Management, Tianjin University of Technology, Tianjin 300384, China.
Computational Intelligence and Neuroscience
|April 30, 2021
Summary
This study introduces a novel DWT-LSTM model for accurate urban nitrogen dioxide (NO2) prediction in Tianjin. The model enhances prediction accuracy by decomposing data, outperforming other methods.
Area of Science:
- Environmental Science
- Data Science
- Atmospheric Chemistry
Background:
- Accurate urban nitrogen dioxide (NO2) concentration prediction is crucial for effective air pollution control.
- Existing models may face challenges in capturing complex temporal patterns of NO2 levels.
Purpose of the Study:
- To develop and evaluate a novel prediction model for daily average NO2 concentration in Tianjin.
- To enhance the accuracy and generalization ability of NO2 concentration forecasting.
Main Methods:
- A hybrid Discrete Wavelet Transform (DWT) and Long- and Short-Term Memory (LSTM) network (DWT-LSTM) model was developed.
- Input data included five major atmospheric pollutants, key meteorological data, and historical NO2 concentrations.
- Data decomposition using DWT increased data dimensionality before LSTM network processing.
Main Results:
- The DWT-LSTM model demonstrated improved accuracy and generalization ability in predicting NO2 concentrations.
- Performance was evaluated against Support Vector Regression (SVR), Gated Regression Unit (GRU), and single LSTM models using Mean Absolute Percentage Error (MAPE).
- The DWT-LSTM model proved more suitable for Tianjin's NO2 prediction compared to the other methods.
Conclusions:
- Decomposing input data into multiple components using DWT enhances data mining capabilities for NO2 prediction.
- The DWT-LSTM model offers a superior approach for forecasting urban NO2 concentrations, contributing to better air quality management.
