Density-based Unsupervised Learning Approach for Generators Coherency Evaluation in Complex Domain

Document Type : Article

Authors

1 Department of Electrical Engineering, Shahid Bahonar University of Kerman, Kerman, Iran

2 Department of Electrical Engineering, Faculty of Engineering, Jahrom University, Jahrom, Iran

Abstract

In measurement-based approaches, generators coherency can be determined on the basis of excited modes, which are obtained in the form of a vector of complex values. In order to find coherent generators in frequency domain, which is suitable for applications such as wide-area control of power systems, these complex vectors should be clustered on the basis of their similarities. To do that, in this paper a new density-based unsupervised learning approach is proposed for clustering complex vectors, which is crucial for finding coherent generators in online applications. The proposed coherency evaluation approach is very simple and more practical according to the reality of power systems since (i) eliminating the complexities in previous studies, it evaluates the coherency from the excited modes point of view using the complex correlation of frequency spectrums, (ii) it uses a new density-based learning approach with only one parameter setting making it more suitable for clustering generators. These features have been demonstrated on the well-known 16-machine, 68-bus system.

Keywords

Main Subjects

Scientia Iranica

Articles in Press, Accepted Manuscript
Available Online from 21 January 2024

Files

Share

How to cite

Statistics