Akademik Veri Yönetim Sistemi
SUSTAINABLE ENERGY GRIDS & NETWORKS, cilt.27, 2021 (SCI-Expanded)
Identification of load in a grid is an essential part of the management of grid and security. In this study, a novel load identification model, a combination of a feature selection method and a classifier, is proposed. Firstly, the features to be used for identification are extracted from current and voltage signals. In this scope 1st, 3rd, 5th, 7th, 9th harmonic currents, the angles of these signals, and THD information are evaluated. ReliefF method is used to reveal each feature's effect on load identification, and then the effectiveness of the features chosen by this method is investigated. A model is then proposed using Radial Basis Function neural network (RBF) and Elman neural networks with selected features, and their performances are compared. The experimental results showed that more successful results were obtained using fewer data thanks to the combined model. Over 95% identification accuracy is achieved with the proposed model. Furthermore, it is shown that the importance of a feature can change depending on the loads in the dataset; therefore, incorporating a feature selection method to classification schema will improve the identification performance. (C) 2021 Elsevier Ltd. All rights reserved.