Statistical and neural network analysis of dissolved gases in power transformers

The onset of electrical discharges or thermal stresses in mineral oil or cellulose insulation of a power transformer can cause the degradation of these materials with the formation of various dissolved gases. These dissolved gases can be extracted and identified with the application of gas chromatography. The overall process, from oil sampling to gas identification, is known as dissolved gas analysis (DGA). In this paper, comparison of conventional DGA interpretation schemes is briefly presented. Moreover, some new artificial intelligence (AI) techniques for transformer incipient fault diagnosis based on DGA data, are also discussed. The second part of the paper reports on the initial work performed for the proposed new approach. This includes simple statistical analysis on DGA records and is followed by high-level data-mining (DM) using self-organising map (SOM) algorithm. The inherent data 'structure' revealed from the latter part of the analysis could hypothetically be associated with certain transformer faults, either electrical, thermal or cellulose decomposition. The proposed approach could provide a viable alternative for transformer incipient fault diagnosis and condition-monitoring applications.