Static security assessment of power system using Kohonen neural network

Abstract

Static security assessment of power systems is a time-intensive task involving repetitive solutions of power flow equations. The issue addressed in this paper is how to substantially reduce the amount of offline security assessment simulations used for neural net training. A Kohonen-based classifier is developed for this purpose. With the proposed scheme, the status of the system security is not needed for all training patterns. Only a selected sample of the training patterns needs to be assessed through simulations. Once the network is adequately trained, neurons that respond to secure or insecure states are self organized in clusters. In the testing stage, the pattern security states is determined by correlating the test pattern with a cluster of a known security status. The proposed scheme also provides information on the degree of system insecurity, and the range of the operation violation.<>