In order to provide a technical support for the optimal operation of the Kaidu River Reservoir and the scientific regulation of water resources, the MIKE coupling model is used to forecast the medium-short-term runoff in the basin. The weather forecast model from the European Centre for Medium-range Weather Forecasts (ECMWF) is selected as the meteorological data input to construct a MIKE SHE model consisting of six sub-modules, namely, evapotranspiration, unsaturated zone, saturated zone, slope flow, river and lake, and snowmelt module. Then, the MIKE Hydro river model which is used for the river confluence calculation is coupled with MIKE SHE model to predict the runoff information in the next ten days. The correlation coefficient (R2), Nash-Sutcliffe efficiency coefficient (NSE) and relative error (BAIS) are used to evaluate the calibration and verification effect of the model, and the prediction efficiency (E) is introduced to evaluate the model prediction accuracy. The results show that in daily runoff simulation of the Kaidu River, the R2 and NSE in the verification period are both above 0.70, and the BAIS is only -15%, indicating a good fit and correlation between the simulated runoff and the measured runoff. In the forecast period, NSE=0.53, R2=0.61, E=0.51, indicating a correlation of 0.61 between the simulated and measured value. The medium-short-term runoff forecast based on the MIKE coupling model can provide a reference for the optimal utilization of water resources in the Kaidu River Basin.