A novel intelligent stability assessment scheme is presented for stability-related feature extraction and weak links identification. An improved genetic algorithm（GA） combined with the k -NN clustering is proposed to select stability-related kernel features from steady-state operation quantities. Simulation test results in the New England 10-machine 39-bus power system support the effectiveness of the proposed algorithm and analysis shows that the selected operation features contain clear information about the instability modes for given faults. By designing the BP-type ANN based CCT estimators with the selected features as inputs， the effectiveness of feature selection is further verified. In the last part of the paper， a feature-overlapping judging scheme is proposed based on an analysis of the correlation between the kernel-feature combination and the corresponding fault locations. The feature-overlapping matrix obtained provides a simple and effective way to realize the stability-based power grid division.