Research on water environmental indicators prediction method based on EEMD decomposition with CNN-BiLSTM

Area of Science:

• Environmental Science
• Data Science
• Water Resource Management

Background:

• Water resource protection is crucial for socio-economic development.
• Traditional models struggle with the complex, nonlinear nature of water quality data.
• Accurate prediction of water environmental indicators is essential for effective management.

Purpose of the Study:

• To develop a novel hybrid model for enhanced water quality index prediction.
• To address the limitations of traditional models in handling seasonal, periodic, and uncertain water quality data.
• To improve the accuracy of predicting key water quality indicators like dissolved oxygen.

Main Methods:

• Ensemble Empirical Mode Decomposition (EEMD) for signal processing.
• Convolutional Neural Network (CNN) for feature extraction.
• Bidirectional Long Short-Term Memory (BiLSTM) network for time-series prediction.
• Hybrid model integrating EEMD, CNN, and BiLSTM for water quality forecasting.

Main Results:

• The proposed hybrid model demonstrated superior performance in predicting dissolved oxygen.
• Significant improvements were observed in prediction accuracy across different time horizons (4-hour, 1-day, 2-day).
• The model outperformed suboptimal models, showing a 5-7% increase in the prediction index.

Conclusions:

• The EEMD-CNN-BiLSTM hybrid model offers a robust solution for water quality prediction.
• This approach effectively captures the complex dynamics of water quality indicators.
• The findings support the application of advanced hybrid models for effective water resource protection and management.