Three typical real-time correction methods, including the traditional error autoregressive method, the nonparametric correction method based on the K-nearest neighbor(KNN), and the multi-cross section correction method based on Kalman filtering, as well as a combination of the KNN method and the Kalman filtering method were used in combination with a one-dimensional hydrodynamic model for flood forecasting and real-time correction in the Wujiadu-Xiaoliuxiang reach of the Huaihe Basin. The principles and construction methods of the four methods are briefly introduced, and a comparative analysis of their simulation results, especially of the stability of the flood peak, the time of occurrence of the flood peak, and the error of occurrence of the flood peak, is conducted. The conclusions are as follows: the three basic methods can improve the accuracy of flood forecasting over quite a long forecast period, and the combination method is more effective and reliable for flood peak simulation, and is more applicable to forecast correction.