A Hybrid Model for Runoff Prediction Using Variational Mode Decomposition and Artificial Neural Network

Abstract

Hydrological runoff prediction in a reliable and precise manner contributes significantly to the optimal management of hydropower resources. Considering the importance of runoff prediction, this study proposed a hybrid model, namely VBH (VMD-BP), coupling variational mode decomposition (VMD) technique, and backpropagation (BP) based artificial neural network (ANN), to predict the monthly runoff of Fentang reservoir, China. Two hybrid models, including ensemble empirical mode decomposition-BP (EEMD-BP) and empirical mode decomposition-BP (EMD-BP), and a standalone BP model, were also developed for comparative analysis. The VBH model performed better compared to the EEMD-BP model in reducing mean absolute error (MAE) by 40.263%, root mean square error (RMSE) by 33.634%, and mean absolute percentage error (MAPE) by 52.906%. The improved results for the VBH model compared to the EMD-BP model included 103.716, 82.266, and 158.303% reductions in MAE, RMSE, and MAPE, respectively. The error reductions by the VBH model compared to the BP model were 113.848% for MAE, 122.022% for RMSE, and 143.026% for MAPE. The results highlighted that the proposed model was superior to the hybrid and standalone counterparts for the hydrological runoff prediction. Water resources designers and planners for future planning and management of hydrological assets can exploit the proposed model.