Research on Monthly Runoff Prediction based on VMD-BP model

Accurately predicting monthly runoff is crucial for studying hydrology and adjusting ecological resources. However, the complexity of monthly runoff data is nonlinear and unstable, which poses significant challenges to improving the accuracy of runoff prediction. This study attempts to use multiple interpretation algorithms and combination models for data preprocessing for comparative experimental analysis. The data preprocessing of this study used VMD algorithm and VMD CEEMDAN algorithm; the prediction model in this article uses BP model and RMF model. Firstly, this article uses the VMD algorithm to decompose monthly runoff data into multiple intrinsic mode functions (IMF), and uses sample entropy (SE) to determine the optimal number of decomposition times. Then, a fully integrated empirical mode decomposition (CEEMDAN) with adaptive noise is used to further decompose the residual components after decomposition. Finally, the BP model and RMF model are predicted by using all decomposed subsequences as inputs. The results show that the VMD algorithm is superior to the VMD-CEEMDAN algorithm, the VMD-BP model is superior to the VMD-RMF model, and the VMD-CEEMDAN-RMF model is superior to the VMD-CEEMDAN-BP model. This article will continue to conduct in-depth research on data algorithm models and prediction models in the future, as well as explore the application of AI intelligence in hydrological prediction models.