面向类别不平衡数据的氧化锌避雷器状态辨识

Study on Status Identification of Zinc Oxide Arrester for Class-unbalanced Data

内蒙古电网的氧化锌避雷器大多位于环境复杂的变电站,承担着重要的过电压保护任务,保障其安全运行对供电可靠性意义重大。避雷器发生故障时,其泄漏电流阻性分量显著增大,测量阻性电流是分析避雷器状态的常用手段。考虑实际运行的避雷器,其状态受多种因素影响,需提取特殊环境下的状态关键特征。由于避雷器状态类分布是不确定的,分析的样本具有明显不平衡性,采用加权极限学习机模型进行分类,可减少因不平衡数据引起的误差,通过附加临时权重增强少样本类的影响,减少多样本类的影响,提高避雷器状态辨识准确率。分析现场试验检测的历史数据,验证了所提方法的有效性,为避雷器状态智能分析提供了一定参考。

Since most of the zinc oxide arresters operating in the Inner Mongolia Power Grid are located in substations with complex environments and undertake important task of overvoltage protection,ensuring their safe operation is of great significance to the reliability of power supply.When the arrester fails,the resistive component of its leakage current increases significantly and it is a common method to analyze the performance of the arrester based on the measurement of resistive current.Considering that the arrester in actual operation is affected by many factors,it is necessary to extract the key characteristics of the state in a special environment.Since the arrester state class distribution is uncertain,and the analyzed samples are obviously unbalanced,it is proposed to use a weighted extreme learning machine model to classify the arrester state to reduce the error caused by unbalanced data.And by adding temporary weights to enhance the influence of small-sample classes,the impact of multiple classes is reduced to improve the state identification accuracy of arrester.The historical field test data are analyzed to verify the effectiveness of the proposed method,which provides certain reference for the intelligent analysis of the arrester state.