Monitoring and assessment of voltage stability margins using artificial neural networks with a reduced input set

A new methodology is proposed for the online monitoring and assessment of voltage stability margins, using artificial neural networks with a reduced input data set from the power system. Within the framework of this methodology, first the system model is reduced using self-organised artificial neural networks and an extended AESOPS algorithm. Then supervised learning of multilayered artificial neural networks is carried out on the basis of this reduced model. Finally, based on the trained network and the reduced set of system variables, monitoring is carried out along with the assessment of voltage stability margins. This methodology is tested comparatively with a methodology for monitoring and assessing voltage stability using a complete input data set. The tests were carried out on a real power system with 92 buses. The results obtained indicate the justifiability of using a reduced system because of the increased efficiency and accuracy of calculation, both in the learning stage and in the recall stage of the artificial neural network.