Towards a Comprehensive DGA Health Index

Dissolved Gas Analysis (DGA) is a widely adopted method for transformer diagnostics and maintenance decision-making. Traditional methods for DGA estimation classify the transformer health state according to predefined gas ratios and intervals. However, these methods are unable to deal with conflicting outputs because their diagnostics output is a single value. There are data-driven DGA classification methods which overcome the limitations of traditional methods, but usually, the DGA result is not considered in isolation and the analysis of other complementary variables can enhance the transformer health state estimation process. Accordingly, this paper presents a novel DGA-based health index formulation combining data-driven models with expert knowledge. Results confirm that th e proposed approach is effective for classifying transformer faults and for identifying incipient abnormal patterns.