Abstract: To obtain relative historical data and to meet the demand of discovering residential customers behavior rule when first designing time-of-use (TOU) price, a model of simulating the demand response behavior rule of residents under TOU price is proposed. Under a certain behavioral assumption of random distribution, the frequencies and probabilities of choosing TOU prices under a variety of given situations for residents in target area are counted through questionnaire designing and random sampling firstly. Secondly, the input and output attributes of the model and training set are designed and determined in line with the questionnaire categories and statistical results. In the end, the response behavior prediction model is built by using least square support vector machine (LSSVM) regression algorithm based on information of the training set. This model can predict the average response results and standard deviations under any TOU price within a certain range to describe residential customers behavior rule under TOU price when there is no historical data, which provides data support for researches in future. Accuracy, rationality and feasibility of the model are verified through case simulation.