Alocação ótima de medidores para fins de detecção de falhas

Abstract

The proposal of this work is to develop a method based on genetic algorithms for optimal allocation of measurements units the distribution grid, which aims to detect monophasic defects with impedance in the medium voltage of distribution systems. The detection is performed by artificial neural networks. The developed methodology for optimal distribution of meters indicates as output the positions that the units should be allocated, creating detection areas. Artificial neural networks were used to perform fault detection in distribution systems, which is an internal routine of the measurement units distribution method. The inputs of the artificial neural networks are voltage and current. These electrical parameters can be either phasor measurements and/or non-phasor, depending on the category of meters defined in the allocation algorithm. The performance of different types of measurement equipment was executed by the analysis of different measurement types. It was selected single-phase defects due to the fact that these are the vast majority of faults in distribution systems. The developed method is able to identify single-phase defects with arc resistance ranging from 0 to 200 ohms, and it is also capable of correctly detect faults when considering load variation. The method of measurement unit allocation and the fault detection algorithm were developed in the software MatLab®. Tests were performed in IEEE systems, presenting good results.