Power system transientstability assessment (TSA) based on patternrecognition techniques can usually be treated as a twopattern classification problem. The classification quality is practically determined by separability of the input space represented by the input features. This paper presents an approach for separability assessment of the input spaces for TSA

through recognizing the “inconsistent cases” of any given sample set. The inconsistent cases in this paper denote the samples that lie in the regions occupied dominantly by the other class in an input space. Through defining a group of hypercubic regions

the inconsistent cases can be automatically sorted out by a process of data mining. A separability index defined based on the amount of the inconsistent cases in the sample set encapsulates the knowledge required for enhancing stability classification. The numerical results for the 10unit 39bus New England power system show the effectiveness of the proposed approach in separability assessment of twopattern input spaces for TSA. The proposed approach is applicable for featureselection procedure and improving classifier design.