Conditional probability‐based interpretation of dissolved gas analysis for transformer incipient faults

In this study, a conditional probability scheme is designed to diagnose transformer incipient faults based on the percentage of five dissolved gases (hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4), and acetylene (C2H2)) with respect to their summation. Based on the fault features, the point probabilities of each fault type occurrence and non-occurrence are computed. Then, the conditional probability of certain fault occurrence is estimated to specify the probabilistic indication of each fault type occurrence. Multivariate normal probability density function is considered to point out the conditional probability of the fault occurrence. It is implemented in different scenarios to define the fault type (partial discharge, low energy discharge, high energy discharge, low thermal, medium thermal, and high thermal) where the best scenario is selected. The proposed technique has the merits of simplicity and ease of implementation. It is assessed through the analysis of 403 dissolved gas sample dataset collected from the Egyptian electric utility as well as from credited literatures. The proposed probabilistic technique is evaluated in comparison with other methods in literatures. Also, it is validated against uncertainty in DGA data up to 20%. The results reveal the ability and reliability of the proposed technique for transformer fault diagnosis.