Detecting the Faults of Subsea Power Cables of Wind Farms with Boosting Ensemble Methods

Eze, Onyedikachi, Guile, Geoffrey and Wang, Wenjia (2022) Detecting the Faults of Subsea Power Cables of Wind Farms with Boosting Ensemble Methods. In: Proceedings of the International Conference on Electrical, Computer, Communications and Mechatronics Engineering (ICECCME). ICECCME, pp. 2492-2497.

[thumbnail of DetectingPowerCableFaults_v4a]
Preview
PDF (DetectingPowerCableFaults_v4a) - Accepted Version
Download (823kB) | Preview

Abstract

The power transmission system is very critical to the functionality and efficiency of offshore wind farms. The system uses subsea cables spanning from deep offshore to the shores for effective integration with other power sources within the transmission and distribution grid. These cables operate under harsh environmental conditions and, as a result, are susceptible to failures. With over 80% of insurance claims so far in offshore wind energy sector, subsea cable failure has huge economic implications. These failures occur as result of fault development within the subsea transmission cable network. This research aimed to develop a supervised machine learning approach to identify and predict these fault developments because if a fault development can be predicted at the incipient stage, planned maintenance or proactive measures can be carried out to avoid degeneration into failure. This paper describes our earlier experiments in applying the extreme gradient boosting ensemble, Gaussian Naive Bayes and decision tree algorithms in solving this problem. The testing results showed that the ensemble algorithms performed accurately and consistently in classifying the faulty cables with an average classification accuracy over 90%.

Item Type: Book Section
Faculty \ School: Faculty of Science > School of Computing Sciences
UEA Research Groups: Faculty of Science > Research Groups > Data Science and Statistics
Depositing User: LivePure Connector
Date Deposited: 15 Dec 2022 01:02
Last Modified: 15 Dec 2022 01:02
URI: https://ueaeprints.uea.ac.uk/id/eprint/90014
DOI:

Actions (login required)

View Item View Item