Daily Runoff Prediction Based on the GWO-PE-VMD-ResNet Combined Model
The traditional decomposition-integration runoff prediction models has shortcomings of low prediction effi-ciency and neglecting component prediction errors.This study proposed a GWO-PE-VMD-ResNet coupled model for dai-ly runoff prediction based on Grey Wolf Optimization(GWO),Permutation Entropy(PE),Variable Mode Decomposi-tion(VMD),and Residual Neural Network(ResNet).Firstly,the GWO algorithm,with permutation entropy as the fit-ness function,was utilized to search for the parameters of the VMD,reducing the uncertainty associated with manual pa-rameter selection.Secondly,the VMD algorithm with selected parameters was applied to decompose daily runoff data in-to several components,reducing the complexity of runoff sequences.Finally,a ResNet runoff prediction model was es-tablished by concatenating and adjusting the runoff sequence components to match the input dimensions of the ResNet model,enabling the prediction of future runoff.This study focused on hydrological stations such as the Yangtze River at Zhutuo,Jianli,and Luoshan for runoff prediction.The results indicate a clear advantage of the proposed model in both prediction accuracy and efficiency.
grey wolf optimization algorithmvariational mode decompositionresidual networkdaily runoff predic-tion
NETL
NSTL
万方数据
