The electricity information acquisition system is prone to errors in the relationship between households in the stations. Traditional diagnostic techniques are mainly aimed at abnormal users in the few stations, but for hundreds of users, there is a difficult problem of extracting the characteristics of abnormal users in multiple adjacent stations. This paper firstly reduces dimension through the principal component analysis of GIS system for area total table and voltage meter data, sets up improved K-means clustering to extract voltage data characteristics, the improved Pearson correlation coefficient algorithm is proposed to analyze the users to be detected, accordingly, the abnormal diagnosis method of household variable relationship based on improved K-means clustering and Pearson correlation coefficient is established to realize the correct diagnosis for multiple abnormal users. The analysis results of practical examples show that the algorithm proposed in this paper can effectively realize the accurate detection and analysis of abnormal users in the case of identifying one or more abnormal users in the same station and multiple abnormal users in different stations. Compared with the traditional detection method, the implementation is simple and more accurate.