Exact Model Determination of a Cable with Variation of Languor and Frequency using Neural Networks

Abstract

A method is developed using a technique based on artificial neural networks for the optimal number of circuit sections (RC, RLC,…) in cascade needed to represent a power cable at selected frequencies for desired levels of precision. As the model of the power cable distributed parameters considered in theory as a reference, the present approach is based on a mathematical analysis to develop a recursive formula, which depends on an infinite cascade cells giving the gain response. By comparing of the distributed parameter model and the recursive formula, the identification of the appropriate number of cells representing the cable at defined frequencies has been determined; this procedure allows to build a sufficient computed database for learning ANN. The simulated results with ANN parameter variations of the environment compared to numerical methods and validated experimentally showed the same efficiency and robustness of our method.