Insulating fault diagnosis of XLPE power cables using multi-parameter based on artificial neural networks

An online monitoring system of XLPE power cables was introduced in the research at first. It could detect the parameters, including partial discharge, dielectric loss, and central insulation resistance and sheathing resistance. The BP artificial neural networks were applied to diagnose the insulating status of XLPE cables using the 16 parameters. The adopted transfer functions in the neural networks were hyperbolic tangent function and S-type function. In order to reduce the training time, the Levenberg-Marquardt training method was used. The experimental results showed that the BP artificial neural networks could be applied in fault diagnosis of XLPE power cables using multiparameter and when the number of nerve unit in the implied layer was fourteen, the output error was least.