Accurate fault location and faulted section determination based on deep learning for a parallel‐compensated three‐terminal transmission line. Issue 13 (31st May 2019)

• Record Type: Journal Article
• Authors: Mirzaei, Mahdi; Vahidi, Behrooz; Hosseinian, Seyed Hossein
• Is Part Of: IET generation, transmission & distribution. Volume 13:Issue 13(2019)
• Page Start: 2770
• Page End: 2778
• Publication Date: 2019-05-31
• Subjects: relay protection -- flexible AC transmission systems -- wavelet transforms -- fault location -- power transmission lines -- neural nets -- power transmission protection -- power transmission faults -- learning (artificial intelligence) -- power transmission control -- power transmission reliability; faulted section determination -- three‐terminal transmission line -- parallel flexible AC transmission systems -- power transmission system -- parallel‐compensated lines -- deep neural network -- parallel FACTS device -- line model -- fault occurrence -- faulted line section -- fault distance -- unsymmetrical fault types -- fault inception angles -- multiterminal lines -- fault resistance -- intelligence‐based approach -- phasor‐based fault location schemes -- SIMULINK -- wavelet transform; Electric power production -- Periodicals; Electric power transmission -- Periodicals; Electric power distribution -- Periodicals; 621.3105
• DOI: 10.1049/iet-gtd.2018.6982
• Languages: English
• ISSNs: 1751-8687

Abstract

Abstract : Parallel flexible AC transmission systems (FACTS) devices affect the performance of protection relays and conventional phasor‐based fault location schemes in transmission lines. This study focuses on both multi‐terminal and parallel‐compensated lines, not investigated simultaneously in previous works. An algorithm based on deep neural networks is proposed for fault location in a three‐terminal transmission line with the presence of parallel FACTS device. The line model and fault occurrence are simulated in SIMULINK and features are extracted from voltages at the three terminals by wavelet transform. The generated features are used to train a deep neural network which determines faulted line section and fault distance simultaneously. The adopted intelligence‐based approach has the advantage of not requiring pre‐knowledge of line specifications, FACTS devices modelling and the uncertainty in compensator parameters. A large number of fault scenarios are investigated. The faulted section is recognised correctly in 100% of test cases. The algorithm performance is acceptable for both symmetrical and unsymmetrical fault types, small fault inception angles and high fault resistance. The accuracy of fault location is improved compared to previous schemes (total mean error of 0.0993%). The proposed algorithm provides an accurate, fast and robust tool for fault location in parallel‐compensated three‐terminal transmission lines.