It is difficult for traditional methods to deal with the mass power user data and spatial correlation characteristics. In order to solve this problem, this project intends to mine and mine the risk characteristics from the power consumer data, and make a deep research on it theoretically. On this basis, a method of power user data clustering based on density clustering is proposed and applied to power system. On this basis, a power user group model based on statistical, temporal and historical characteristics is proposed. Through the test of experimental data, the prediction accuracy of the model established in this paper reached 89%–97%. On this basis, a new risk assessment method for power customer data is proposed.