An intelligent analysis method of large customers energy consumption based on stream computing is designed to comprehensively analyze their power and energy consumption and deal with the inability of finding abnormal power consumption in time. First, the overall structure of the intelligent analysis system for the energy consumption of large customers is given, and the requirements for software and hardware functions are described. Relying on the real-time processing capabilities of stream computing technology and high-performance data throughput capabilities to process the collected data in real time, machine learning is used to learn to process abnormal data in accordance with the user's comprehensive energy consumption analysis mathematical model and electricity theft identification model. It also displays the line loss calculation curve, energy consumption curve, voltage, current and other electrical data within a certain period of high-risk users to provide auxiliary support for inspection work. This work is supported by the Science and Technology Project of South China Power Grid (No. 059300HK42180085).