Association rule analysis on confidence of features for transformer faults

作者：Zheng Yuan Bing^*; Sun Cai Xin; Li Jian; Chen Wei Gen; Wang You Yuan

来源：High Voltage Engineering, 2012, 38(1): 82-88.

DOI：10.3969/j.issn.1003-6520.2012.01.012

• Apriori
• Apriori algorithms
• ChiMerge algorithm
• Confidence
• Confidence values
• Continuous data
• Data mining algorithm
• Diagnosis algorithms
• Discretization method
• Discretizations
• Fault types
• Multi-value
• On-line monitoring system
• On-line monitors
• Online monitoring
• Transformer faults

As the potential information of every feature from on-line monitoring system of transformers is not the same amount for different fault types of transformers, it can be helpful for fault diagnosis and prediction of transformers to quantify the relationship between a specific type of fault and each feature. In this paper, the continuous data discretization of on-line monitoring of transformer is performed by Boolean-value discretization method and multi-value discretization method based on ChiMerge algorithm respectively. Then the improved Apriori association rule data mining algorithm is used to calculate the confidence value between the features of on-line monitor system of transformers and the several of fault types. Finally, in the instance the calculation of the confidence between a number of features and a several of fault types of transformers is conducted, and the result is shown that the different relationship exists between a specific feature and a several of fault types of transformers. The results of examples also show that it is significant to improve the efficiency of diagnosis algorithm by quantifying the relationship. Additionally, the multi-valued association rule mining is carried through in the examples, and results show that association rules can be applied for fault diagnosis of transformers when the fault types of transformers are divided specifically.

• 出版日期2012
• 单位输配电装备及系统安全与新技术国家重点实验室