Skip to main content
placeholder image

Evaluating the effectiveness of a machine learning approach based on response time and reliability for islanding detection of distributed generation

Journal Article


Download full-text (Open Access)

Abstract


  • Conventional relays, such as vector surge relay, frequency relay and rate-of-change-of-frequency relay, are usually employed for islanding detection; however, these conventional relays fail to detect islanding incidents in the presence of small power imbalance inside the islanded system. This study presents an islanding detection approach for synchronous type distributed generation using multiple features extracted from network variables and a support vector machine (SVM) classifier. Features are extracted from a sliding temporal window, whose width is selected so as to achieve the highest detection rate at a fixed false alarm rate. The SVM classifier is trained with linear, polynomial and Gaussian radial basis function kernels, and the parameters of the kernels are tuned to improve the classification performance. The application of the proposed method is illustrated for islanding cases associated with different power imbalance conditions, including small power imbalance conditions associated with the non-detection zone of conventional relays. Furthermore, variation of detection time as a function of power imbalance scenarios, which involve all probable combinations of deficit of active/reactive and excess of active/reactive power imbalance, is assessed in the testing phase. The performance of the proposed approach is evaluated and compared with those of conventional relays in terms of reliability and response time of islanding detection.

Publication Date


  • 2017

Citation


  • M. Rezaul. Alam, K. M. Muttaqi & A. Bouzerdoum, "Evaluating the effectiveness of a machine learning approach based on response time and reliability for islanding detection of distributed generation," IET Renewable Power Generation, vol. 11, (11) pp. 1392-1400, 2017.

Scopus Eid


  • 2-s2.0-85030094992

Ro Full-text Url


  • http://ro.uow.edu.au/cgi/viewcontent.cgi?article=1738&context=eispapers1

Ro Metadata Url


  • http://ro.uow.edu.au/eispapers1/737

Number Of Pages


  • 8

Start Page


  • 1392

End Page


  • 1400

Volume


  • 11

Issue


  • 11

Place Of Publication


  • United Kingdom

Abstract


  • Conventional relays, such as vector surge relay, frequency relay and rate-of-change-of-frequency relay, are usually employed for islanding detection; however, these conventional relays fail to detect islanding incidents in the presence of small power imbalance inside the islanded system. This study presents an islanding detection approach for synchronous type distributed generation using multiple features extracted from network variables and a support vector machine (SVM) classifier. Features are extracted from a sliding temporal window, whose width is selected so as to achieve the highest detection rate at a fixed false alarm rate. The SVM classifier is trained with linear, polynomial and Gaussian radial basis function kernels, and the parameters of the kernels are tuned to improve the classification performance. The application of the proposed method is illustrated for islanding cases associated with different power imbalance conditions, including small power imbalance conditions associated with the non-detection zone of conventional relays. Furthermore, variation of detection time as a function of power imbalance scenarios, which involve all probable combinations of deficit of active/reactive and excess of active/reactive power imbalance, is assessed in the testing phase. The performance of the proposed approach is evaluated and compared with those of conventional relays in terms of reliability and response time of islanding detection.

Publication Date


  • 2017

Citation


  • M. Rezaul. Alam, K. M. Muttaqi & A. Bouzerdoum, "Evaluating the effectiveness of a machine learning approach based on response time and reliability for islanding detection of distributed generation," IET Renewable Power Generation, vol. 11, (11) pp. 1392-1400, 2017.

Scopus Eid


  • 2-s2.0-85030094992

Ro Full-text Url


  • http://ro.uow.edu.au/cgi/viewcontent.cgi?article=1738&context=eispapers1

Ro Metadata Url


  • http://ro.uow.edu.au/eispapers1/737

Number Of Pages


  • 8

Start Page


  • 1392

End Page


  • 1400

Volume


  • 11

Issue


  • 11

Place Of Publication


  • United Kingdom