Toggle Main Menu Toggle Search

Open Access padlockePrints

Anomaly Warning and Fault Detection in DC Pico-grid with enhanced k-Nearest Neighbours Technique

Lookup NU author(s): Yang Quek, Dr Wai Lok Woo, Dr Thillainathan Logenthiran


Full text for this publication is not currently held within this repository. Alternative links are provided below where available.


© 2018 IEEE. The k-nearest neighbours (kNN) algorithm, which is usually used for classification, is presented in this paper to detect faults and trigger anomaly warnings in a single sensor multiple loads dc pico-grid. Anomalies warning is getting more attention in the recent years as it can used as a trigger for predictive maintenance, which is preferred over repair work after a fault detection. On top of performing its usual duty of load classification in the circuit during normal operation, the kNN algorithm is enhanced with 3 additional techniques to set 3 anomaly criteria for the triggering of alarm when the extracted features of the test object exhibit abnormal behaviours. The experiment is set in a dc pico-grid as there is a growing interest and demand in dc loads. Experiments with various anomalies show that the proposed enhanced algorithm can effectively detect anomalies and faults.

Publication metadata

Author(s): Quek YT, Woo WL, Logenthiran T

Publication type: Conference Proceedings (inc. Abstract)

Publication status: Published

Conference Name: International Conference on Innovative Smart Grid Technologies, ISGT Asia 2018

Year of Conference: 2018

Pages: 728-733

Online publication date: 20/09/2018

Acceptance date: 22/05/2018

Publisher: IEEE


DOI: 10.1109/ISGT-Asia.2018.8467961

Library holdings: Search Newcastle University Library for this item

ISBN: 9781538642917