The invention discloses a weighted K-means method for grouping of power customer values. The invention uses a weighted K-means clustering algorithm which is suitable for the characteristic variable data features of power customers. The algorithm comprises steps of first determining that the sum of standard deviations of the data in a weighted power customer group is used as a clustering criterion function, wherein the weight is a ratio of the number of customers in the power customer group to a total number of customers; then when calculating the similarity between the power customer objects and a customer group center point according to the criterion function, by using Euclidean distance as a premise, using a data standard deviation in the customer group as a weight of a reference factor to achieve more accurate value grouping of nonuniform-density power customer objects . A grouping clustering result of the weighted K-means clustering algorithm applied to the grouping of power customer values shows that the method is suitable for practical operation data and has an effect of improving clustering compactness. More high-quality grouping clustering results can also ensure efficient decision implementation, and ultimately bring higher profits to the power supply enterprises.