摘要
在气体绝缘组合电器(gas insulated switchgear,GIS)实体模型中分别放置了针-板、悬浮金属颗粒和绝缘子表面固定金属颗粒放电模型,用超声波传感器采集到其放电波形。对放电波形提取的特征向量进行局部线性嵌入(local linear embedding,LLE)算法降维处理,用降维后的向量作为输入对BP_Adaboost分类器进行训练和测试类型识别。识别结果表明,用这样方法进行GIS绝缘缺陷类型识别可以在减少计算量的同时保持较高的识别率,说明了其在局部放电模式识别应用中的有效性。
Needle -plate, suspended metal particles and metal particles fixed on insulating surface discharge models are placed in gas insulated switchgear （GIS） entity model.The corresponding discharge waveforms are collected by the ultrasonic sensor.The eigenvectors of the discharge waveforms has been processed by local linear embedding algorithm to reduce the dimension.The processed vectors are used as the input to train and test BP _ Adaboost classifi-er.Recognition results show that GIS insulating defects recognition with this method will reduce the calculation amount and keep a high recognition rate at the same time , which will illustrate the effectiveness in the application of partial discharge pattern recognition.
出处
《电测与仪表》
北大核心
2014年第15期37-41,共5页
Electrical Measurement & Instrumentation
基金
国家高技术研究发展计划项目(863计划)(2011AA05A121)
中央高校基本科研业务费专项资金资助项目(13ZD14)
