This paper selects two small watersheds in the middle reaches of the Yellow River and uses the similar watershed method to explore the applicability of the distributed hydrological model SWAT to simulate runoff in a small watershed with no data in the middle Yellow River. Using the methods of similar distance, similar watershed area and similar properties, the known watershed rate parameters were transplanted to the no data watershed for runoff simulation. The model was determined by the daily runoff and the monthly runoff from 2009 to 2013. The efficiency coefficient (NS) and the decision coefficient (R2) were used as the evaluation indicators. The sensitivity coefficients of the four models were determined, and the daily runoff from 2014 to 2016 was used to validate the model. The results show that the regular monthly runoff simulation R2 is 0.76 and NS is 0.70. For the daily runoff simulation, R2 is 0.70, NS is 0.64. For tthe validation period, R2 is 0.82, NS is 0.74 for the monthly runoff simulation, and R2 is 0.78, NS is 0.68 for the daily runoff simulation. The SWAT model based on the similar watershed method has a good applicability to the monthly runoff simulation of small watersheds with no data.