Application of fuzzy equivalent matrix for fault diagnosis of oil-immersed insulation

For power equipment, the relation between its faulty phenomena and the corresponding faulty reasons is somewhat fuzzy. By using the data of dissolved gas analysis (DGA), the fuzzy cluster analysis (FCA) technique was applied to identify the fault patterns of oil-immersed power transformers and to overcome the defects of conventional expert systems (ES). As a kind of typical FCA method, the fuzzy equivalent matrix (FEM) algorithm was used. After some fuzzy mathematical processing from the original DGA data, a FEM could be obtained. By clustering the matrix under different fuzzy similar numbers, the clustering type sets were acquired. There must be a suitable set responsible for the final categories of faults. Compared with the conventional methods, it isn´t based on inferring rules, but abstracting fault information from the original data. It reveals the practical advantages of unsupervised systems, including the ability to produce categories without supervision. According to the clustering results, the new method can identify the different insulation states, such as normal state, arc discharge or overheating, etc. and enhance the reliability of decisions. Conversely, some of the DGA data will be mis-diagnosed if only the IEC approach is used. Thus the diagnostic conclusions obtained from the new method will be more reliable