Investigation of artificial neural networks for voltage stability assessment

This paper investigates the use of artificial neural networks for determining the voltage stability limit of a power system during contingencies. Different neural network architectures are trained with data containing a variety of load patterns, generation patterns, generator voltage pattern and transformer tap ratio settings, via economic minimization. Tests are conducted on a neural network selected based on training performance. Studies are conducted on the New England 39 bus system. The work is an enhancement of previous work by the authors and provides criteria for possible extension of the stability limit. It is concluded that the selected neural network architecture gives reasonably accurate predictions of the collapse point with potential for improving the stability margin