Load classification is of great importance in load modeling. It is well known that it's difficult to judge the rationality of the type of a load substation through real time data and to check the validity and accuracy of the clustering results for lack of a checking method. The daily load data from EMS/SCADA is employed and its relative parameters are chosen as eigenvectors to classify the substation. A hypothesis proof-test based on formula T and Monte Carlo method are presented to guide the fuzzy clustering process and inspect the correctness of the clustering results. A case study shows the validity of the novel method through its application in a local power system.