HYBRID SUPERVISED AND UNSUPERVISED NEURAL-NETWORK APPROACH TO VOLTAGESTABILITY ANALYSIS

This paper presents a hybrid neural network approach to voltage stabil ity analysis. Essentially, the proposed hybrid neural network consists of a Kohonen network and a multi-layer feed-forward network. By using voltage collapse margin method (VCM) and singular value decomposition method (SVD), the hybrid neural network is trained to identify voltag e weak buses/areas and to evaluate the loadability of power systems in terms of voltage stability. Moreover, the Kohonen neural network work s as a front end to cluster input patterns with similar features of op erating conditions and hence the generalization capability of the mult i-layer feed-forward network has been improved significantly. The effe ctiveness of the proposed network has been demonstrated on the IEEE 57 -bus test system and the results are very encouraging. (C) 1998 Elsevi er Science S.A. All rights reserved.