High-frequency ANN-based Method for Fault Detection, Classification and Location on Multi-terminal HVDC Links

Abstract

This work is the investigation of novel protection schemes for the direct current (DC) link of VSC-based high voltage direct current (HVDC) and multi-terminal high voltage direct current (MTDC) systems, with the purpose of isolating the faulted transmission line rather than tripping the entire network when faults occur on the transmission line. The main essence of the developed method is using the fast Fourier transform (FFT) feeding a multi-layer perceptron (MLP) artificial neural network (ANN) to harness short circuit fault current frequency components for MTDC overhead line relaying. This approach is shown to give more reliable fault detection, classification and location algorithms that are not affected by fault resistance. Detection by analysis of a short circuit fault current only meaning that no voltage signal is required. The protection algorithms are based on the relationship between the magnitudes or phases of fault current frequency component with fault.

A system consisting of two-terminal VSC converters and a three-terminal HVDC system with equivalent grid interconnection and source representation is built using PSCAD/EMTDC software with a detailed switching model for the converter components, which are used for obtaining the fault current data for transmission line terminals. The method is verified by studying different cases with a range of fault resistances in various fault locations, and in addition, external faults. The results show that the proposed method gives fast and reliable fault detection and classification and accurate location for DC line faults.

The method is shown to accurately detect, classify and locate overhead line faults. Unlike existing travelling wave-based methods, which must capture the initial wavefront and require high sampling rates up to 192 kHz, the new approach is more robust since it gives accurate fault detection and fault location over a range of windowed post-fault signals. Furthermore, the proposed method is fault resistance independent meaning even a very high fault impedance has no effect on the accurate fault location.

Date of Award	24 May 2017
Original language	English
Awarding Institution	University of Bath
Supervisor	Simon Le Blond (Supervisor)