Applications of artificial intelligence and hybrid neural network methods with new bonding method to prevent electroshock risk and insulation faults in high-voltage underground cable lines
Küçük Resim Yok
Tarih
2018
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
SPRINGER LONDON LTD
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
The sheath current causes cable faults and electroshocks in the high-voltage underground cable lines. Thus, the sheath current must be determined before a new high-voltage cable line is installed to take the required precautions against the sheath current effects. In this study, PSCAD/EMTDC is used for simulation of high-voltage underground cable line to examine the sheath current. The sheath current is forecasted with artificial neural network (ANN) by using the results of simulation studies, and differential evolution algorithm (DEA), genetic algorithm (GA), clonal selection algorithm (CSA), hybrid DEA-CSA, artificial bee colony (ABC), particle swarm optimization (PSO) and inertia weight PSO (iPSO) are used to find the best weights of ANN. The sheath current formation factors which are obtained in PSCAD/EMTDC are used as inputs of ANN. It is seen at the end of simulation studies that many factors affect the formation of the sheath current, so Pearson correlation is used to determine the most effective sheath current formation factors for inputs of ANN. In the literature, three types of bonding methods are used to reduce the sheath current and voltage, but these methods are not sufficient to reduce both sheath current and the sheath voltage simultaneously. Therefore, in this study, the sectional solid bonding method is developed as new bonding method to reduce the sheath voltage and current in this study. When the sectional solid bonding is used, the sheath voltage can be reduced under touch voltage, and the sheath current can be reduced under the desired value.
Açıklama
Anahtar Kelimeler
Artificial intelligence, Neural network, High-voltage underground cable, Sheath current
Kaynak
NEURAL COMPUTING & APPLICATIONS
WoS Q Değeri
Q1
Scopus Q Değeri
Q1
Cilt
29
Sayı
8
