FRA winding type field cases and automatic classification using AI
Power transformers are critical to all High Voltage (HV) networks and are quite reliable over the long term, with a typical failure rate of around 1 % or less.
byJean Sanchez, David Gopp
Using machine learning to validate and identify SFRA characteristics
Power transformers are critical to all High Voltage (HV) networks and are quite reliable over the long term, with a typical failure rate of around 1 % or less [1]. The CIGRE Working Group (WG) A2.49 has elaborated that offline as well as online monitoring and operational data are crucial factors for condition-based transformer assessment [2]. The quality of the data plays a viable role so that the right conclusions can be drawn from it. This includes both the acquisition and the processing of data. The SFRA method is one of several offline methods which is commonly used and has shown to be the most sensitive and non-invasive method to detect mechanical, especially electrical, faults. More specifically, the mechanical integrity of the core, windings, and clamping structure as well as the electrical integrity, such as shorted windings and turns. [3] To improve its diagnosis, artificial intelligence tools are tried to be used, like interesting proposals of [4] §6.4, to automate at best diagnosis firstly and somehow better understand winding designs, mostly unknown to the final clients, and then its faults in future works. This paper tries to assess winding design through FRA field cases by knowing some of it from manufacturers and to give new ways of automatically assessing FRA curves from the field as often difficult, especially for non-experts.
Power transformers’ active part is made of windings, manually manufactured for all ratings above a few MVA. The design challenges on windings are to carry the current and to withstand the high voltage and its transients. Two main winding designs are now used in the world, core type and shell “pancake” type designs (under Westinghouse license, historically), with significantly different geometry. Both types are working well, even if the core type is more widely manufactured in the world than the shell type, and both are subject to the same challenges.
To read the article, subscribe and choose the option which suits you best. We offer both free and paid options, and the registration takes only a minute.
Subscribe to Transformers Magazine
