The customer power load has the characteristics of complex and irregular use. Therefore, in order to meet the demand response scheduling requirements, cluster analysis should be carried out according to the user load characteristics. To solve the above problems, this paper proposes a multi-type demand response user portrait research method based on improved k-means clustering algorithm. Firstly, the improved k-means clustering algorithm is used to analyze the user's electricity consumption data, study the user's electricity consumption behavior, and classify the user. Then, considering the dynamic elastic adjustment coefficient, the price-based demand response model is constructed. Finally, select the demand response characteristic index, construct the demand response characteristic evaluation system and establish the user demand response portrait. The validity of the model is verified by analyzing the energy use data of a park in Shandong Province. The example analysis shows that the method used in this paper can effectively reflect the acceptability of demand response of different types of users under different electricity prices.