The abnormal behavior of users' electricity consumption does not know how to affect the access equipment and users, and may even endanger the healthy operation of the power grid. Therefore, the analysis of the abnormal behavior of electricity consumption is very important. Based on big data and machine learning technology, a power consumption abnormal behavior analysis system is designed, and proposes the overall framework and related configuration of the system design. The design system can perform abnormal analysis on the electricity consumption, voltage quality, load and three-phase unbalance rate, reactive power and power factor of the user's electricity, and display it to the administrator and users in a visual way. At the same time, it conducts early warning and tracking treatment for high-risk users, and investigates and analyzes the behavior of electricity theft. This system effectively analyzes the user's abnormal electricity consumption behavior and can carry out early warning of electricity theft, which plays a certain role in the good operation of the power grid.