State of The Art Machine Learning based DGA Analytical Solution

Abstract

Power Transformers is one of the most important and critical equipment in power system. Accordingly, transformer health condition is an essential component in the overall system resilience. The healthiness of power transformer is assessed through several means, among which is the Dissolved Gas Analysis (DGA). The accurate assessment of the DGA result is a key factor of identifying the transformer condition. Currently, DGA is mainly analyzed by the subject matter experts capitalizing on their experience or industrial guidelines such as IEEE. The main challenge is that the DGA assessment is an art that varies from transformer to transformer based on many factors which might impact the accuracy of the assessment. This challenge is aggravated with speared of online DGA tools for power transformers requiring immediate automated assessment. In this paper a state-of-the-art approach is used to classify transformers condition utilizing a supervised classification machine learning model that analyzes the transformers DGA data and serves as an effective decision-making basis regarding transformers condition.