Based on the historical data of power customers, the index system needed to determine the model is determined based on the customer's basic attributes, power usage behavior, payment behavior, customer credit, geographic region, and weather environment. Through the correlation coefficient matrix and information value (IV) of the indicator, the index variables that finally enter the model are selected, the variables are grouped by the optimal grouping method and the weight of Evidence (WOE) is transformed. Based on the processed data, the logistic regression algorithm is used to construct the PV poverty alleviation user analysis model, and the users are classified into high, medium and low grade suspected poor users according to the analysis model, and the analysis results are fed back to the government's poverty alleviation office for confirmation feedback. This paper provides a basis for selecting poor households in the Xinjiang Poverty Alleviation Competition.