Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
University of Strathclyde
Online conditional anomaly detection in multivariate data for transformer monitoring
A common problem with retrofitting sensors to an aged transformer is that it displays signature behaviour which is imperfect, but which represents the normal operating condition. This research presents a unique approach to characterising normal behaviour for individual transformers, accounting for environmental and operational factors. This approach can reduce false-positives, where an alarm is raised where no fault exists, thus saving utilities needless maintenance and wasted engineer time. A direct output of EPSRC-funded research [EP/D034531/1, £2.5M, 2006-2010], this research was collaborative with National Grid [Graham.Moss@nationalgrid.com], via a 1-year case study.