邻域粗糙集与相关向量机相结合的变压器故障综合诊断模型

Comprehensive Fault Diagnosis Model of Transformers Based on Combination of Neighborhood Rough Set and Relevance Vector Machine

针对变压器相关向量机故障诊断模型中冗余信息影响泛化能力的问题,提出了邻域粗糙集与相关向量机相结合的变压器故障综合诊断模型。首先采用领域知识和快速约简算法进行属性约简;其次利用条件属性对决策属性的依赖性度量进行属性加权;然后将约简后和数值化后形成的特征向量集输入相关向量机进行训练;最后用测试集进行测试。实例显示所提方法的测试确诊率均高于单独相关向量机模型,说明邻域粗糙集提升了相关向量机的实用性和准确性。

To deal with the issue of generalization ability affected by redundant information in the relevance vector ma- chine （RVM）based fault diagnosis model of transformers, this paper proposes a comprehensive fault diagnosis model based on the combination of neighborhood rough set （NRS） and RVM. First, neighborhood information and quick re- duction algorithm are employed to reduce the attribute reduction. Then, the dependence of conditional attribute on deci- sion attribute is used to acquire the attribute weight. Next, the feature vector set obtained after reduction and numeral- ization is input into the RVM for training. Finally, tests are conducted with test set. A case study shows that the diagno- sis rate with the proposed method is higher than the RVM model, which further indicates that NRS enhances the practi- cability and accuracy of RVM.