By using the daily electricity data and meteorological data from 2009 to 2010 in Wuhan, an algorithm is presented based on the load decompositon, in which the daily maximum load is divided into the fundamental load and the weathersensitive load related to the weathersensitive load index that integrates the synthetically influence of temperature, relative humidity, and wind speed.By using the method of correlation ratio, the nonlinear relation between weathersensitive load and sensitive load index is analyzed， and the changing ratio between weathersensitive load index and load is analyzed. Based on the least meansquared deviation and nonlinear generalized least square, the polynomial forecasting model for weathersensitive load in Wuhan power grid in summer is established. Application tests show that the mean predicting errors are about 6 percent in summer from 2011 to 2012. Thus, the model can be used in the meteorological service in power grid dispatching in summer.