A clustering method for large user load based on the characteristics of electricity is proposed.Then the demand response model of the large user is divided into the load reference component part and the load response component part.A short term load forecasting method based on multivariate time series reconstruction is used to get the user’s load reference component.On the basis of the demand price theory, the load probability model of the single large user response TOU price is obtained by the response of the actual data.Matlab is used to simulate the actual data of Nanjing area.The example shows that the model can well reflect the situation of the large user’s response for TOU price.which provides scientific basis formulating strategy for TOU price.