Abnormal behavior of electricity consumption not only affects access equipment and users themselves, but also endangers the normal operation of power grid. Therefore, the analysis of abnormal behavior of electricity consumption is very important. Based on big data and machine learning technology, an analysis system of abnormal behavior of electricity consumption is designed, and the overall framework and relevant configuration of the system design are proposed. The designed system can analyze the abnormal factors such as power consumption, voltage quality, load, three-phase unbalance rate, reactive power and power factor, etc, and show them to administrators and users in a visual way. Meanwhile, early warning and tracking processing are conducted to high-risk users to carry out investigation and analysis of stealing electricity. The system can effectively analyze abnormal behaviors of users and give early warning of electricity theft, which plays a key role in the stable operation of power grid.