Pattern recognition techniques applied to the classification of swing curves generated in a power system transient stability study

This paper presents two approaches to determine the stability of power system based on pattern recognition techniques using artificial neural network (ANN) and linear classification. The two major states of power system operations are termed stable and unstable. The performance index can be expressed by the patterns and then be recognized by a properly trained neural network or a linear discriminant function. A feature vector selected by fast Fourier transformation is employed for reducing input pattern dimension. ANN is found to be an efficient tool for identifying stable states. System stability or instability indices can be predicted quickly and accurately.