According to the characteristics of different correlation and importance between the variables of transformer state evaluation, the weighted Mahalanobis distance is introduced into the traditional multivariate statistical control chart based on Mahalanobis distance, and a transformer evaluation method based on the variable weighted multivariate statistical process control is designed. This method combines the weight of transformer state variable given by association rule method, and uses the wave detection ability of multivariate control chart to transform the multivariate real -time monitoring data corresponding to various faults of transformer into a test statistic, and then evaluates the operation state of transformer. The example analysis of the current circuit overheat fault shows that compared with the current transformer condition evaluation standard, the method has faster fault detection speed; the retrospective detection of the sample data of known faults shows that the method has higher fault detection rate.