摘要
为了提高对变压器罕见故障的诊断准确率,有效应对变压器油中溶解气体(DGA)数据存在的样本不均衡现象,提出了一种基于加权极限学习机(weighted extreme learning machine,WELM)的变压器故障诊断方法。研究了加权极限学习机的参数对分类准确率的影响,明确了参数设置过程中的主要参数和次要参数;在此基础上提出了一种WELM的参数选择方法;给出了基于WELM的变压器故障诊断的基本流程与具体方法。实验结果表明,加权极限学习机在变压器故障诊断中具有易用性和有效性,并对少数类样本有更高的识别准确率。
To improve the diagnosis accuracy of rare faults of transformers, and to make effective response to the data imbalance in dissolved gas analysis （DGA） data, a transformer fault diagnose method based on weighted extreme learning machine （WELM） is presented. Firstly, the study of how the parameters of WELM affect its performance is done to specify the primary parameter and minor parameter in the configuration progress. On this basis a method of selecting the WELM parameters is proposed. The procedure and steps of transformer fault diagnosis using weighted extreme learning machine is provided. The experimental results show the ease of use and the effectiveness of WELM applied to power transformer fault diagnosis, and it enhances the accuracy in identifying the minority sample class.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第12期4340-4344,共5页
Computer Engineering and Design
基金
河北省自然科学基金项目(E2009001392)
