The management of station line loss is the core issue of consumption side operation and management of power grid company, and it is also the basis of improving quality and efficiency to guarantee power grid profit. Relying on the power information collection terminal, the power grid marketing system collects and maintains a large number of user data. Using the real time collected power grid big data to detect the operation of the power grid and carry out timely maintenance is an effective way to control the line loss. Although there are a lot of data collected by power grid monitoring, it is still insufficient to face the complicated causes of line loss disqualification. Using the known collection data, establishing the causal link between data and fault causes can greatly reduce the work pressure of operation and maintenance personnel. In the process of data acquisition, communication and maintenance, there must be inaccuracy in massive monitoring terminals. Using rough set theory to solve the uncertainty of user side data is proposed,and the characteristic data under different faults are given. The internal relationship between faults and data are analyzed, and the decoupling of fault parameter characteristic data is realized. The classification of fault causes in the station area is given, and the correctness of the proposed classification method is verified.