基于SMOTE和决策树算法的电力变压器状态评估知识获取方法

提出基于合成少数过采样技术(SMOTE)算法和决策树算法的电力变压器状态评估知识获取方法,首先针对变压器非正常状态样本数量较少的情况,采用SMOTE算法补充非正常状态样本数量,解决了变压器样本集类别不平衡问题。然后将变压器状态评估过程视为分类过程,利用决策树模型为白箱模型的特点,将变压器状态评估知识获取问题转化为构建决策树的问题。最后采用C4.5决策树算法构建决策树,从中提取变压器状态评估知识,得到关键变压器状态量和评估规则。以某地市级供电公司110 kV电压等级油浸式变压器实际数据开展实例分析,结果表明所提出的方法能实现状态评估知识的自动化获取,可以为该地区110 kV油浸式变压器的状态评估工作提供决策支持。

Knowledge acquisition method of power transformer condition assessment based on SMOTE and decision tree algorithm

A knowledge acquisition method of power transformer condition assessment based on SMOTE(Synthetic Minority Over-sampling TEchnique) algorithm and decision tree algorithm is proposed. Firstly, for the case that samples in abnormal condition is much fewer than samples in normal condition, SMOTE algorithm is used to supplement the number of abnormal state samples to solve the imbalance problem of the transformer sample set. Then the transformer condition assessment is treated as a classification process, and the knowledge acquisition problem is transformed into a decision tree building problem using the characteristics of decision tree model as a white box model. C4.5 algorithm is applied to construct decision tree, then the transformer condition assessment knowledge and rules are derived from the decision tree. Case study is performed using 110 kV transformer data collected from a distribution company. Study results indicate that the proposed method can automatically acquire knowledge and rules and can provide decision support for future transformer condition assessment work.