Abstract: In the context of future high penetration of new energy, the uncertainty of supply-demand balance gradually increases. Demand response is an important means of ensuring the balance of power and electricity in the system by tapping into user-side flexible resources. When power sector works on demand response, historical data is needed for an initial assessment of load response potential, so as to select the users with high potential and initiate mobilization efforts. This article focuses on defining and providing a mathematical expression for load step that represents the energy consumption characteristics of industrial users. And then a user selection method for industrial demand response based on load step is proposed. Firstly, an index system for the potential of industrial users' demand response across multiple time scales based on load step is proposed. And then, a user selection model is established to conduct an initial evaluation of different users' response potential, and the k-means algorithm and the nearest neighbor propagation algorithm are used to divide groups, allowing for user selection across different time scales. Finally, a case study is presented based on actual load data from several industrial users in industries such as cement and paper, illustrating the user selection results for industrial demand response using the proposed method.