Aiming at the evaporation prediction problem of water body covered by floating balls, this study introduces a correction term g(m) based on the Priestley-Taylor model. By analyzing the relationship between coverage and latent heat flux, the exponential ( high temperature season ) and linear ( transition season ) correction forms are determined, and robust regression method is used to reduce the interference of outliers. The results of cross-validation showed that the accuracy of the exponential model was significantly better than that of the linear model ( D = 0.986, NSE = 0.947, RMSE = 0.54 mm·d^-1 ), and the parameters had strong stability and no systematic deviation. In the verification of high coverage in different places, the model also showed good adaptability ( NSE = 0.734, D = 0.890 ). The improved model proposed in this paper constructs a unified prediction framework for different float coverage rates based on basic meteorological parameters, realizes accurate estimation of evaporation process under different coverage rates, and provides an efficient and practical calculation method for water resources management in arid areas.