In order to solve the evaluation problem of multi-level flood control operation schemes under different flood control level objectives, taking the the Taihu Lake Basin as study area, a comprehensive evaluation index system of multi-level flood control and waterlogging drainage operation schemes for basins, regions and cities was constructed. And a comprehensive evaluation model of operation schemes was constructed based on K-nearest neighbor (KNN) and random forest (RF) algorithms. The results show that the combination evaluation of KNN-RF is achieved by combining KNN model and RF model. The average relative error and average absolute error of KNN-RF combining model in the operation scheme evaluation for the three levels of flood control objectives, namely, basin, region and town, are reduced to 1.25%, 0.82%, 2.43% and 0.511,0.342,1.38, respectively. The maximum relative error and maximum absolute error have been improved, and the overall accuracy rate of grading is higher than 95%. The KNN-RF combining model can screen out optimal operation schemes for various levels of flood control objectives, reduce abnormal evaluation errors caused by the uncertainty of a single algorithm, and significantly improve the evaluation accuracy.