Abstract:The management of line loss refinement has always been a key work of the national power network company. In recent years, the technology of data acquisition for line loss has become more and more mature. However, the analysis method is still lacking, which has seriously affected the effect of high loss management. In order to solve the problem of poor management of high losses in power grid operation, we propose a method of large data analysis to solve the problem of traditional line loss management. Firstly, a three-layer linear damage anomaly analysis model composed of Hampel resistance algorithm, weighted Pearson algorithm and random forest algorithm is built. Through this model, we can combine the characteristics of different users ''different power consumption, and analyze the users of large power anomaly and small power anomaly from top to bottom. Then the model is embedded in the existing line loss monitoring platform of the company, which can realize real-time monitoring, effective mining, depth analysis, precise positioning and work order control of the line loss data in the province, and form a new method of line loss anomaly analysis and processing based on the three-layer analysis model. This method is effective in analyzing and treating 10kV high loss lines and 0.4 kV high loss platform areas.