An Improved Method for Fault Diagnosis of Oil-Immersed Transformers Based on Simulation Test Platform
Lin Yang, Lingzhi Gao, Xuesong Luo, Yanpeng Hao, Zhijun Zhang, Ying Jin, Jiantao Zhang
Abstract
Existing methods for the IEC ratio method have problems such as codes absence and absolute boundaries. To solve the above problems, a test platform for simulating faults in transformer oil is built in this article. The simulation tests are conducted on thermal, electrical, and electro-thermal faults of needle-plate and column-plate electrodes under oil and oil-paper insulation. 398 groups of gas chromatography data are obtained and 20 groups of these data are selected to validate the test platform. Among them, the fault types of 37 groups of gas chromatography data cannot be diagnosed or are incorrectly diagnosed. These groups are categorized into two datasets with codes absence and absolute boundaries. Based on the analysis of them, an improved three-ratio method is proposed. The fault criteria for codes absence “000” and “011” are supplemented. Then, fuzzy logic is used to diagnose the fault types for the data with absolute boundaries. A test set including 113 groups of samples is constructed. It is composed of samples from the 37 groups mentioned earlier, IEC Technical Committees, and relevant literature. The comparison with effective fault diagnosis methods shows that the highest accuracy of diagnosis for methods such as Doernenburg, Rogers, three-ratio method, and other ratio methods with good diagnostic effectiveness mentioned in other literature does not exceed 63%. In contrast, the diagnostic accuracy of fault types and faults involving insulation structure of the mentioned improved method is 81.42% and 81.13%, respectively, which is a better improvement effect.