A FUNCTIONAL LINK APPROACH TO NEURAL NET COMPUTING OF CRITICAL FAULT CLEARING TIME

Abstract

The aim of this study is to synthesize the critical clearing time (CCT) of a power system through successive pattern space -transformations using aritificial neual networks. The critical clearing time is an attribute which provides significant information about the quality of post fault system behaviour. It represents a complex mapping of the pr•e_-fault, on-fault and post fault system conditions into time domain. A feed forward neural network has beent trained to learn. this mapping and successfully perform under variable system operating conditions and topologies. But the supervised method of learning of input/output pairs through back propagation of error leads to a slow convergence to the desired solution. A modification of this method has been attempted to simplify the network structure. This modified system is able to discover what combination of measurements are significant in determining CCT-and quick screen power system contingencies. Besides simplifying the measurement requirements, this system offers an improved convergence time.