There is a difficulty in identifying station line loss rate anomalies and tracking associated users accurately. Thus a method of station line loss rate anomaly research and tracking associated users based on multi-dimensional scene division is proposed. First, a data preprocessing strategy based on the improved complete EEMD with adaptive noise (ICEEMDAN) algorithm is proposed. By analyzing the changing trend of the historical line loss rate curve of the station area, the extensive scene set of the station line loss rate is constructed. On this basis, the SVM clustering algorithm is used to make further quadratic partition to scene set, thus to establish multi-dimensional scene set. Secondly, an interval dynamic translation strategy based on the number of cluster cases is proposed to determine the interval range of the line loss rate standard library in different scenarios. Also an interval overlap rate strategy is used to merge the isolated intervals in the division, so as to realize the complete division of the line loss rate standard library in the station area. Finally, an abnormal line loss tracking method based on correlation analysis is given. This determines the strong correlation factors of abnormal line loss through grey correlation analysis, and quantitatively analyzes the internal correlation degree between each factor and station users based on an improved Adtributor algorithm to improve the tracking accuracy of abnormal line loss rate users. Cases are simulated and analyzed by using real data in station area. The simulation results show that the proposed method is effective and practical.