Monthly runoff prediction of the lower Yellow River based on the VMD-SSA-KELM model

Abstract

Guiding the efficient utilization of water resources relies on a high-precision prediction of monthly runoff in the lower reaches of the Yellow River. disaster prevention and mitigation, water environmental protection, and ecological restoration. Based on the superior performance of VMD in processing non-stationary monthly runoff sequences, the multimodal optimization ability of SSA in the direction of data sequences, and the advantageous features of KELM model KELM model efficiency, tuning free, and memory friendliness, a monthly runoff prediction model for the lower Yellow River was established using the VMD-SSA-KELM coupling method, and apply the coupled model to predict the monthly runoff at Lijin Hydrological Station in Downstream Areas. The results indicate that the model has a reasonable predictive effect on the monthly runoff data of this hydrological station; it has high accuracy compared with the traditional prediction model, and the R2 of the prediction model for Lijin Hydrological Station reaches 0.97, with an average absolute error of 8.02, an average absolute percentage error of 0.44, and a root mean square error of 37.25; at the same time, the model can effectively extract the inherent feature information of the corresponding time series, improving the prediction performance of runoff data, it can make the monthly runoff forecast more accurate.