Application of Discrete Wavelet Transform Method To The Analysis of A Submarine Cable Trip

Correct and rapid fault clearing as well as minimizing power outage durations are important utility company concerns. Protections should work properly to isolate faults and quickly to minimize system impacts while analysis should facilitate timely restoration. This paper examines tripping of an undersea cable by a numerical line differential (87L) relay. While conventional DFR data analysis was helpful for determining where the fault was not, it was unable to definitively determine analysis where the fault was, including which end of the zone was faulted. This paper explores the use of discrete wavelet transform methods to better determine the fault nature and location.

The cable in question is a 30-mile 46kV undersea cable running from the town of Harwich, Massachusetts to the island of Nantucket serving as one of two power supplies from the mainland. The complete differential zone of protection includes a section of overhead bus on the mainland as well as the cable termination inside of gas-insulated switchgear (GIS) on the island. A mainland side DFR record showed a low magnitude phase A to B fault with a half-cycle duration prior to the 87L relay operation which corresponded with AB fault targets from the 87L relay.

Whereas a 30-mile overhead line can be readily surveyed by arial patrol, it’s almost impossible to perform a detailed inspection of the same length of underground and undersea cable within a relatively short time. The DFR record indicating a phase-to-phase event with no ground involvement suggested the underground and undersea cable portions of the protection zones could be ruled out but that is not a definitive finding. With current magnitude being very low and the duration being only ½ cycle, the DFR record also did not definitively establish that this was even a fault event. An analysis of SF6 gas sampled from the GIS revealed water and arc products which suggested possible dielectric breakdown of the gas. Unfortunately, all this evidence was suggestive but not definitive.

This lack of definitiveness prompted interest in exploring discrete wavelet transform (DWT) methods to improve the determination of this incipient fault’s nature and location from the DFR data. DWT methods provide a microscope for closer inspection of digital signals and processes and have been widely used in many industries and in artificial intelligence applications.

DWT tools within PSCAD/EMTP were chosen to analyze COMTADE formatted files from the DFR record. Faulted phase currents were decomposed under different frequency spectrums in the time domain. Waveforms under certain frequency ranges can be clearly observed and harmonic signatures can be determined. DWT analysis and removal of the fundamental frequency uncovered the presence inter-harmonic currents, consistent with arc discharges in gas as opposed to a solid dielectric material. After reviewing this paper, the reader will have an introduction to applying wavelet transform principles to fault analysis.