Comparative study of rainfall prediction based on different decomposition methods of VMD

Area of Science:

• Hydrology and Water Resources
• Artificial Intelligence
• Signal Processing

Background:

• Effective rainfall forecasting is crucial for water resource management and disaster mitigation.
• Traditional forecasting methods often face challenges in accuracy and adaptability.
• Variational Mode Decomposition (VMD) offers potential for signal analysis in complex time series.

Purpose of the Study:

• To evaluate the effectiveness of different Variational Mode Decomposition (VMD) strategies for rainfall prediction.
• To develop and test an improved forecasting model integrating VMD with advanced optimization and neural network techniques.
• To compare the performance of the proposed model against traditional methods using real-world data.

Main Methods:

• Variational Mode Decomposition (VMD) with full and stepwise decomposition approaches.
• Modified African Vultures Optimization Algorithm (MAVOA) enhanced with Tent chaotic mapping.
• Differentiable Neural Computer (DNC) integrating recurrent neural networks and computational processing.
• Comparative analysis of forecasting models using data from four sites in the Huaihe River Basin.

Main Results:

• The Single-model Fully stepwise decomposition (SMFSD) VMD approach combined with MAVOA-DNC demonstrated superior performance.
• The SMFSD-MAVOA-DNC model achieved an average Root Mean Square Error (RMSE) of 9.02 and Mean Absolute Error (MAE) of 7.13.
• The model attained an average Nash-Sutcliffe Efficiency (NSE) of 0.94, outperforming traditional VMD full decomposition and other coupled models.

Conclusions:

• Stepwise decomposition of VMD, when integrated with the MAVOA-DNC model, provides a highly effective rainfall forecasting solution.
• The proposed SMFSD-MAVOA-DNC model significantly enhances prediction accuracy compared to conventional methods.
• This advanced forecasting approach offers substantial improvements for water resource management and disaster preparedness.