Study On Diagnosis For Power Transformer Faults Based On Association Rule Mining

The power transformer is a major apparatus in a power system. Early diagnose and detect the incipient faults is necessary for avoiding catastrophic failures, costly outages. The existing methods sometimes fail to determine the fault type, the workers onsite often use several methods or other related information about the transformer to do such judge work. So exploring new way to resolve this early warning problem is still a challenge. Association Rule Mining (ARM) is one of the data mining techniques used to extract hidden knowledge from datasets and there is a close relation between the transformer oil and the transformer fault type. This thesis is using ARM to explore the relation between transformer oil and fault type (state), and obtain the rules for diagnosis of the new transformer fault (state). In this thesis, a transformer fault diagnosis based on ARM method is presented and illustrated detail, a comparison with the IEC three-ratio method is carried to validate its effectiveness. With knowledge base, the implement of transformer fault diagnosis is achieved and introduced. Example and experiment show that it is feasible to use ARM and knowledge base for transformer fault diagnosis.