This paper presents a new K-Medoids clustering algorithm based on gray relational degree. Analyze the gray incidence of each attribute and convert them into the weights of the attributes, and then apply these weights to the distance measure of the cluster; based on this measure, this paper proposed an improved clustering algorithm: Gray-K-Medoids clustering algorithm and applied it to the analysis of the aluminum electrolysis data. The paper introduces the gray relational degree and the basic principle based on the gray relational degree clustering and introduced the improved algorithm in detail. In order to test the effect of improving the algorithm, it was used to the production data of an aluminum plant, and the results show the effectiveness of the algorithm, has a certain promotional value.