Improving forecasting accuracy of medium and long-term runoff using artificial neural network based on EEMD...

Area of Science:

• Hydrology
• Data Science
• Environmental Engineering

Background:

• Hydrological time series forecasting is crucial for effective reservoir management.
• Accurate medium and long-term runoff predictions are essential for water resource planning.
• Traditional forecasting methods may struggle with the complexity of hydrological data.

Purpose of the Study:

• To develop and evaluate an Artificial Neural Network (ANN) model integrated with Ensemble Empirical Mode Decomposition (EEMD) for enhanced runoff time series forecasting.
• To assess the model's performance in medium and long-term forecasting horizons.
• To compare the proposed EEMD-ANN model against a standalone ANN approach.

Main Methods:

• Decomposition of original runoff time series into intrinsic mode functions (IMFs) and a residual series using EEMD.
• Forecasting of each IMF component and the residual series using individual ANN models.
• Ensemble forecasting by summing the predictions of all components to reconstruct the final runoff forecast.

Main Results:

• The EEMD technique effectively decomposed the runoff time series, revealing underlying data characteristics.
• The EEMD-ANN model demonstrated superior performance in medium and long-term runoff forecasting compared to the standard ANN model.
• Performance evaluation using RMSE, MAPE, R, and NSEC metrics confirmed the enhanced accuracy of the proposed method.

Conclusions:

• Ensemble Empirical Mode Decomposition (EEMD) is an effective preprocessing technique for improving hydrological time series forecasting accuracy.
• The integrated EEMD-ANN model offers a significant advancement over conventional ANN methods for medium and long-term runoff prediction.
• The developed model provides a robust tool for reliable reservoir management and water resource planning.