Dynamic voltage stability assessment and control by computational intelligence

This paper presents a novel approach for the assessment of dynamic voltage stability. The power system dynamic voltage stability problem is modeled as a multi-input multi-output (MIMO) transfer function. By means of singular value analysis, interactions between properly defined input and output variables affecting dynamic voltage stability can be analyzed for different frequencies. The output singular vectors provide information about the level of risk at the observed nodes and the input singular vectors show which control variables are most effective for taking counter measures. Since the modal analysis and singular value analysis required for MIMO systems are very time consuming, this paper proposes using computational intelligence (CI) to predict the input and output singular vectors, which are of interest in an on-line power system operation. In this study, a neural network (NN) is used for the prediction of singular vectors. Moreover, in order to reduce the extremely high number of the NN outputs, the principal components (PC), instead of the full singular vectors, are suggested as NN outputs. After calculation of the PC's by the NN, the full vectors can be obtained by a simple matrix multiplication