Substation load consists of a variety of user load

whose characteristics are very complex. Singly choosing daily load curve or the proportion of users as indicators for clustering may cause ignorance of other factors and lead to inaccuracy results. Therefore

in this paper

a multivariate clustering model is proposed

which takes both daily load curve and the users of substations into account. To solve the model

K-means algorithm is used to analyze the daily load curve data

then a two-stage clustering-correction algorithm is put forward to correct the last clustering results according to proportion of users in substations. In conclusion

the clustering result after correction achieves high accuracy

which can be effective to prevent the result from falling into the local optimum and clearly reflect the difference of both load curve and user proportion among substations.