Abstract: The non-stationarity of monthly runoff series in Weihe River Basin is increasing and it is difficult to predict accurately, a multi-step prediction model of monthly runoff series based on optimal variational mode decomposition(OVMD),stochastic configuration networks(SCN) and recursive multi-step prediction strategy was proposed.Firstly, OVMD was used to project the runoff data into subsequences with different frequencies.Then, SCN was used to predict each decomposition part, and the single-step prediction results were obtained by superposition.Finally, the recursive multi-step prediction method was used to predict the runoff data for a long time in the future, and the multi-step prediction results were obtained.The measured monthly runoff time series of Huaxian Hydrological Station and Xianyang Hydrological Station from 1970 to 2019 were selected for case analysis, and we compared the predicted results with other popular models.RMSE,MAE,MAPE and NSE were selected to evaluate the prediction results.The results showed that the NSE of the OVMD-SCN model in the single-step prediction experiments of Huaxian Hydrological Station and Xianyang Hydrological Station reached 98.15% and 98.52%,respectively, which were significantly higher than other popular models.In the multi-step prediction experiments of two hydrological stations, the evaluation indexes of OVMD-SCN were better than other popular models.It showed that the proposed method can accurately predict the runoff in the future 5 months.The research results can provide a new method for monthly runoff prediction in the Weihe River Basin.