Abstract: Power transformers are one of the most important and critical assets in the electricity distribution and transmission network. Power quality (PQ) can be disturbed when a power transformer is brought into or out of service, so it is very important to be sure of the reason for this action. Dissolved gas analysis (DGA) in oil can be used to diagnose the condition of transformer insulation. There may be situations where the DGA results indicate the presence of a serious fault which would lead to the transformer being taken out of service, when in fact the high gas concentrations are due to the leakage into the main oil tank of gases generated in the on-load tap-changer (OLTC) during normal operation. In previous work, using machine learning techniques and a distribution system operator's DGA database, a decision tree (DT) was developed to identify oil contamination from OLTC gases. In this work, the developed DT is applied to a new DGA database to identify contaminated transformers and test its accuracy. A total of 1161 DGA results from 95 transformers with OLTC were used, giving an initial DT accuracy of 83.13% when all samples were analysed and 85.26% when the last DGA result from each transformer was used

Authorship

Download reference