The multivariate statistical method is introduced in the paper to analyze conventional statistical operators of partial discharge（PD）. By using factor analysis， eight new features are extracted from 27 operators that effectively realize the data reduction. These new features not only represent variance information of original operators by a limited number of variables， but also are independent of each other and have their own specific properties. An accelerating electrical-thermal aging test is implemented on an artificial oil-paper model with an internal flat air gap. During the aging test， PD signal acquisition is carried out periodically. By using the new features extracted by the proposed factor analysis method， the clustering and discrimination results of the samples with different aging stages provide significant information in the assessment of oil-paper insulation condition.