Abstract: For the purpose of improving the performance of automatic extracted fine-rule,a method of linear decision tree based on support samples was proposed for rule extraction in this paper.Support samples,which are located in the vicinity of real stability boundary and used as input of decision tree,are selected by distance method.Linear decision tree based on linear classifiers,which were generated by linear transformation,was proposed to obtain integrated attributes operation rules for power system security and stability study.Case study was carried out in one-machine infinite bus system as well as IEEE 39-bus system.Results show that,rules generated by the proposed method are small in structure,have high generation ability and interpretability,and provide sensitivity information,which is valuable for assistant decision-making.Meanwhile,generation time is remarkably reduced.Advantages mentioned above makes that the proposed method is potential in extraction fine security and stability operation rules.