With the gradual advancement of electricity market reform, demand response will play an increasingly important role in future new power systems. Considering the problems of cumbersome process of DR potential calculation and lack of detailed data on user electricity consumption process, a group users DR potential prediction and classification method based on historical load, temperature, and electricity price data is proposed. First, data processing and information extraction are carried out on the user’s daily electricity load curves.Three characteristics, monthly load regularity, daily load fluctuation and peak-valley consistency, are calculated, which form an index system of physical regulative potential evaluation. Further, nonlinear au-to-regressive model with exogenous inputs neural network is applied to the prediction of group users’daily load and DR potential. Finally, taking industrial users as examples, Meanshift algorithm is used to partition user clusters, and DR regulation power of general component manufacturing industry is predicted. By comparing and analyzing with actual data, the effectiveness of the proposed method in this paper is verified.