ARTIFICIAL NEURAL NETWORK MODELS OF CROSS-LINKED POLYETHYLENE

Cross-linked polyethylene (XLPE) is the most widely used insulator material in high-power cables. The complex electrical permittivity of the XLPE layer mostly determines the leakage admittance of the cable and the propagation speed of the signal. The complex electrical permittivity of XLPE depends on not only operating frequency but also temperature. In this study, Artificial neural networks (ANNs) are used to model the complex electrical permittivity parts of the XLPE. The structure of the ANNs is optimized. It has been found that the optimized ANN can predict the behavior of the XLPE with an R2 value of 0.99.