ARTIFICIAL NEURAL NETWORKS FOR PREDICTING THE GASSING TENDENCY UNDER ELECTRICAL DISCHARGE IN INSULATING OIL FOR EXTENDED TIME
Main Article Content
Abstract
The object of this study is to investigate the comportment under electrical discharge of mineral oil and natural ester for extended time, study the effect of electrical discharge on the physical, chemical and electrical properties of both fluids and use an intelligent system for predict the gassing tendency during extended time aging using the most significant parameters in degradation.
Various scenarios were considered. The study was carried on new fluids under electrical stress for different ages. The 6802, 6181 and 924 tests are used in measure of parameters in degradation. All the results obtained are summarized and compared.
For predicting the gassing tendency for a extended ages, and in measure that the high periods can't attended in laboratory; models using the stepwise regression technique are proposed, and implanted in a neural network system, this neuronal model has been tested on laboratory samples obtained in ISOLIME and high prediction accuracy has been achieved.
Various scenarios were considered. The study was carried on new fluids under electrical stress for different ages. The 6802, 6181 and 924 tests are used in measure of parameters in degradation. All the results obtained are summarized and compared.
For predicting the gassing tendency for a extended ages, and in measure that the high periods can't attended in laboratory; models using the stepwise regression technique are proposed, and implanted in a neural network system, this neuronal model has been tested on laboratory samples obtained in ISOLIME and high prediction accuracy has been achieved.