Development of Trend Detection Technique for Dissolved Gas Analysis of Transmission Power Transformers

Abstract

Trend detection in Dissolved Gas Analysis (DGA) data is crucial for diagnosing the health of transformer insulation systems. The complexity of this task arises from gas level fluctuations, varying DGA monitoring frequencies and changes in gas patterns over time. This paper presents a novel automated trend detection technique based on the Mann-Kendall test, tailored for large-scale industrial DGA databases. The technique not only identifies trends but also quantifies the confidence levels of these trends, providing more detailed insights for transformer asset managers. An example application on a substantial DGA database from transmission power transformers, including in-service units and those with dielectric faults, overheating in winding and overheating of elements outside of winding, reveals distinct trend characteristics. The proposed technique serves as an automated asset management tool, facilitating rapid scanning of large DGA databases for improved DGA data interpretation and utilisation.