An advanced event tree analysis method for detecting faults in power transformers

Relevance. is improving the efficiency of decision support for the trouble-free operation of responsible electrical equipment requires improved methods and algorithms for fault detection, which is undoubtedly an urgent task. The purpose. is to develop an effective and reliable procedure for the holistic analysis, assessment and prediction of possible malfunctions in oil-filled transformer equipment, as one of the critical assets of an energy company. Methods . When solving problems, the dissolved gas analysis (DGA) method is used as the most informative in terms of the technical condition operational assessment of the transformer and detection of signs of its deviations from the norm, as well as an improved method for constructing and event tree analysis (ETA), which combines the traditional structure of cause-and-effect relationships with components for calculating the equipment health index. Results. The improvement of the event tree analysis method concerns the introduction of a quantitative measure of the importance of cause-effect relationships at each of the hierarchical levels, with the determination of the probabilities of outcomes of each of the initiating events based on a given root event probability. The role of a quantitative measure is performed by a weighting coefficient system obtained using the matrices of paired comparisons of Saati for each ETA level. A logical and probabilistic ETA model has been formed, which allows not only to record the interrelationships between events and the sequence of their occurrence, but also to rank them according to the degree of influence on the final undesirable outcome. Calculations confirming the proposed approach effectiveness have been performed. Conclusion . A logical-probabilistic ETA model has been obtained, which provides, along with the DGA method, a holistic analysis, assessment and prediction of possible malfunctions of oil-filled transformer equipment. An effective procedure has been developed for constructing and analyzing data sets using expert assessments and components for calculating the equipment health index. The possibility of effective application of the developed model is demonstrated by analyzing the causes of monotonous excess of the boundary concentrations of carbon monoxide and carbon dioxide dissolved in the oil of the studied transformers.