摘要
高压电缆局部放电样本数量对模式识别精度有较大影响,样本数量不足,模型难以得到充分训练,从而影响识别精度,对局部放电样本进行扩充可以有效解决该难题,对此,提出基于可变噪音的高压电缆局部放电样本扩充方法。首先,构建5种高压电缆典型缺陷并进行加压测试,获取局部放电原始数据;然后,介绍基于可变噪音的局部放电样本扩充原理;最后,采用随机森林、支持向量机、逻辑回归3种方法来验证所提方法的可行性。结果表明:基于可变噪音的局部放电样本扩充方法可以提高局部放电模式识别精度;在噪音水平为0.01时,对随机森林模型的识别精度提升了3.15%,提升效果最好。
The number of partial discharge(PD)samples,which is applied for PD pattern recognition of the high voltage cable has a great impact on pattern recognition accuracy.Pattern recognition models are difficult to be fully trained if the number of PD samples is insufficient,thus failing to achieve high recognition accuracy,which could be improved by PD sample augmentation.Therefore,this paper proposes a PD sample augmentation method for high voltage cables base on variable noise method.Firstly,five typical defects of high voltage cables are constructed,and PD samples are obtained by high voltage testing.Then,the theory of variable noise method is introduced in details.Finally,three methods including random forest,support vector machine and linear regression are used to verify the effectiveness of the proposed method.The results show that it is able to improve the pattern recognition accuracy by using the proposed PD sample augmentation method based on the variable noise method.At a noise level of 0.01,the recognition accuracy of the random forest model is improved by 3.15%,which represents the best improvement effect of the proposed method.
作者
刘文浩
李文泽
王洪雨
吴毅江
王干军
Ashfaque Ahmed Bhatti
彭小圣
LIU Wenhao;LI Wenze;WANG Hongyu;WU Yijiang;WANG Ganjun;BHATTI Ahmed Ashfaque;PENG Xiaosheng(Zhongshan Power Supply Bureau of the Guangdong Power Grid Co.,Ltd.,Zhongshan,Guangdong 528400,China;State Key Laboratory of Advanced Electromagnetic Engineering and Technology,Huazhong University of Science and Technology,Wuhan,Hubei 430074,China)
出处
《广东电力》
2020年第9期35-41,共7页
Guangdong Electric Power
基金
国家自然科学基金项目(51541705)
广东电网有限责任公司科技项目(GDKJXM20172769)。
关键词
高压电缆
局部放电
模式识别
可变噪音
数据扩充
high voltage cables
partial discharge
pattern recognition
variable noise
data expansion
作者简介
刘文浩(1986),男,广东高州人,工程师,研究方向为输配电及用电工程,E-mail:55977285@qq.com;通信作者:李文泽(1994),男,内蒙古兴安盟人,在读硕士研究生,主要研究方向为电力设备状态检测及故障诊断、新能源功率预测等,E-mail:1043384339@qq.com;彭小圣(1983),男,湖北随州人,副教授,博士,IEC工作组专家,主要研究方向为电力系统主设备智能监测、局部放电在线监测与模式识别、新能源功率预测等,E-mail:XiaoshengPeng@hust.edu.cn。
