Key feature selection is of great significance for intelligent assessment of transient power angle stability of power system. This paper proposes a key feature selection method for improved Relief. Through analyzing and comparing the applicability of correlation coefficient, mutual information and Relief, three representative key feature selection methods, in the intelligent assessment of transient angle stability of power system, the applicability of Relief is better. The Relief index reflects the strength of the feature to classify transient power angle stability and instability. By eliminating redundancy and increasing the weight of unstable samples, the accuracy of index calculation and the ability to adapt to sample imbalance are improved. Based on improved Relief, feature is selected by iterative optimization, which can automatically screen out the optimal key feature set. The effectiveness of the method is verified by an example of power grid.
 

feature selection;transient power angle stability;improved Relief