Modelagem das curvas tempo x corrente de elos fusíveis do tipo expulsão por meio de redes neurais artificias

Abstract

Protection studies are essential to maintain the levels of energy supply in accordance with standards imposed by regulatory agencies. Currently, this type of study is carried out through computational tools. Therefore, a correct modeling of protection devices is essential. Among the devices used in power distribution systems protection, the most used ones are fuse cutouts. Fuse cutout consists of 3 components: base, fuse holder and fuse link. The fuse link is composed by 3 curves: Minimum Melting (MM), maximum melting and Total Clearing curve (TC). In this way, numerous works model the MM and TC curves of the fuse links through mathematical expressions. Due to the non-linear behavior of the curves, this task becomes complex. In order to overcome this adversity, this dissertation proposes the use of Artificial Neural Networks (ANNs). The results obtained are presented and a comparative analysis with other works is carried out. In addition to RNA, two mathematical functions were evaluated for modeling the TCC curves of the preferred “K” and “H” fuse links, with RNA being the technique that obtained the best results. The MATLAB software was used to develop the methods. To evaluate the models, the IEEE 34 Node test feeder was implemented in the DIgSILENT software. The system was modified for the insertion of fuse cutouts and through the Monte Carlo Method short circuits were applied at the end of each branch. In this way, the operating time of the fuse links was obtained. The operating times show that the proposed methodology provides a satisfactory and promising TCC model for implementation in programs dedicated to protection studies.