CNN–SVM Based Fault Detection, Classification and Location of Multi-terminal VSC–HVDC System

• Published in: Journal of electrical engineering & technology 2023, 18(4), , pp.3335-3347
• Main Authors: Gnanamalar, A. Jasmine, Bhavani, R., Arulini, A. Sheryl, Veerraju, M. Sai
• Format: Article
• Language: English
• Subjects: Electrical Engineering; Electrical Machines and Networks; Electronics and Microelectronics; Engineering; Instrumentation; Original Article; Power Electronics; 전기공학
• ISSN: 1975-0102; 2093-7423
• DOI: 10.1007/s42835-023-01391-5

Offshore wind farms (OWF) are emerging steadily throughout the previous decade due to the steady growth in electricity demand. A high voltage direct current (HVDC) transmission network based on voltage source converters (VSC) is an inexpensive solution for long-distance large power transmission that is suitable to link OWF to an alternating current network. In VSC based multi-terminal HVDC systems, DC fault protection is a considerable difficult. To increase the reliability of system protection, fault detection, classification, and location identifications are essential. This keeps electrical systems functioning properly continuously and reduce economic losses, but it might be difficult. To overcome these challenges, this study proposed a computational efficient integrated convolutional neural network–support vector machine (CNN–SVM) approach model while preserving the accuracy. In this paper, the Hilbert–Huang transform (HHT) is used to extracts a feature from current signals. Next, the proposed SVM–CNN algorithm detects, classifies, and computes the fault's location in multi-terminal VSC-based HVDC systems within 2 ms. The simulations were performed using the MATLAB R2019b software. To improve the reliability of system protection, the performance evaluation of the proposed CNN–SVM model includes a comparison with prior-art techniques such as SVM, ANN, and RNN. The simulation results demonstrate that this proposed strategy can identify, classify and locate the VSC–HVDC transmission system DC faults in multiple fault circumstances with high speed and accuracy. The experimental findings reveal that the proposed approach achieves a better fault classification accuracy range of 99.87% within 2 ms. The accuracy of the proposed technique is 0.03%, 0.022%, and 0.03% of higher than existing ANN, SVM, RNN techniques, respectively.