Back to Search
Start Over
Prediction of insulating transformer oils breakdown voltage considering barrier effect based on artificial neural networks.
- Source :
-
Electrical Engineering . Dec2018, Vol. 100 Issue 4, p2231-2242. 12p. - Publication Year :
- 2018
-
Abstract
- The insulating oil performance could be enhanced in high-voltage apparatus using barriers. The importance of the barrier in increasing the dielectric strength of the insulating oils in order to reduce the oil failure stresses had not been sufficiently studied. In this paper, the effects of the barrier variables on the insulation performance of the transformer oil for point-plate and plate-plate gaps were demonstrated. These variables are: gap space (d), the barrier location relative to the high-voltage electrode (a / d) %, barrier diameter (D), barrier thickness (e), electrode configurations (EC), the presence of contaminating particles, the weight of the contaminating particles (W) and the temperature of the insulating oil (T). The statistical t test was used to explain whether the effect of these parameters was significant or not. Furthermore, the above-mentioned variables were used as training variables to construct the prediction model of oil breakdown voltage considering barrier effect based on the artificial neural networks (ANN). The ANN model was developed based on the results from experimental works. Therefore, 784 samples were used as training data set and other 25 samples were taken as testing and validating samples. The results explained that the prediction ANN model had a high ability to expect the breakdown voltage for other different experiment cases. The average errors of the training and testing samples were 1.6%, and 2.66%, respectively. Therefore, the prediction accuracy could be considered as 98.4% for training and 97.34% for testing. [ABSTRACT FROM AUTHOR]
Details
- Language :
- English
- ISSN :
- 09487921
- Volume :
- 100
- Issue :
- 4
- Database :
- Academic Search Index
- Journal :
- Electrical Engineering
- Publication Type :
- Academic Journal
- Accession number :
- 133032708
- Full Text :
- https://doi.org/10.1007/s00202-018-0697-5