基于模糊关联规则挖掘的电力变压器故障诊断方法

In the process of condition monitoring of transformer, characteristic quantities of a fault are uncertain in some cases. Therefore,the diagnosing process of transformer operation status is somewhat fuzzy,so the definite association rules fail to reflect effectively the relation between the characteristic quantities and the fault. In order to establish a more accurate and effective model for transformer fault diagnosis, we extend the optimized classic Apriori algorithm to the affairs containing fuzzy attributes after using the principal component analysis to optimize the multi-source parameters,and combine it with the IEC three-ratio code as the characteristic quantity to extract rules. The efficiency and accuracy of the established model are higher compared with that of the original Apriori algorithm. The model can also be applied to association rules mining of multiple parameters of the transformer. In conclusion, the common extraction rule of characteristic quantities on the basis of combining Apriori algorithm with IEC 3-ratio code has higher hit ratio, and the characteristic quantity extraction rule combining the fuzzy theory with the IEC 3-ratio code offers higher hit ratio than that rule combining the classic set theory with the IEC 3-ratio code. In summary, the combined extraction rule of characteristic quantities via the Apriori algorithm achieves higher hit ratio than that via IEC 3-ratio code. The present fuzzy association rule model can diagnose transformer faults more accurately and efficiently.