Abstract
Artificial neural network (ANN) algorithms have been applied successfully on very wide range of applications in power systems. In high voltage engineering, ANNs have been applied efficiently and effectively for pattern recognition of partial discharges. A major field of ANN application is function estimation, because the useful properties of ANNs such as adaptivity and nonlinearity are well suited to function estimation tasks where the equation describing the function is unknown as the only prerequisite is a representative sample of the function's behavior. In this paper, the pre-requisite training data are available from experimental studies performed on models of polluted insulators under power frequency voltages representing different pollution levels ranging from light to severe pollutions. Extensive detailed studies and tests have been carried out to determine the ANN parameters to give the best available results and to assess the effect of the presence of inadequate data in the training set on modelling accuracy. The new approach using an ANN as function estimator is employed to model accurately the relationship t=f (V, L, R/sub p/). It is found that, when training is complete, ANN is capable of estimating the flashover time very efficiently and effectively even when the inadequate data are incorporated in the training set. The present study clearly indicates the efficacy of ANN as function estimators in the insulator flashover studies.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have