Comparison between backpropagation and revised GMDH techniques for predicting voltage harmonics

Two multilayered-network-based methods are presented for predicting voltage harmonics in power systems. One method is the backpropagation technique, and the other is revised group method data handling (RGMDH). In order to investigate their effectiveness, the two methods are applied to observed data of voltage harmonics that seem to be randomly varying. A comparison is made between two models in terms of model accuracy. The results show that the backpropagation algorithm gives much better results due to the adaptive model structure