基金项目:
国家自然科学基金项目(51807072);
广东电网有限责任公司科技项目(GDKJXM20172769);
National Natural Science Foundation of China (51807072);
Project Supported by Science and Technology Project of Guangdong Power Grid Co;
, Ltd;
, (GDKJXM20172769);
New feature construction and optimal feature selection of partial discharge (PD) pattern recognition for HV cables contribute not only to improvement of pattern recognition accuracy and efficiency, but also to PD parameter visualization of HV cable condition monitoring and diagnostics. In the paper, a novel random forest (RF) algorithm based optimal feature selection for PD pattern recognition of HV cables is proposed. Based on five types of artificial defects of 11 kV ethylene-propylene (EPR) cables, 3500 sets of transient PD pulses and 3500 sets of typical interference pulses are extracted. 1235 features in total are extracted. Then the RF based optimal feature selection for PD pattern recognition of HV cables is carried out. The feature ranking results of PD signals and interference signals as well as the ranking results of different PD signals are obtained and evaluated with pattern recognition methods based on back propagation neural network (BPNN) and support vector machine (SVM). Results show that, the wavelet combination features and the features describing “fast pulses” and “slow pulses” are effective features for identification of PD and interference signals. The wavelet combination features are attested to be effective for recognition of different types of PD signals. The RF method is proven effective for PD features selection of HV cables and is prospective to be applied to PD feature selection of other HV power apparatus.
KEY WORDS :partial discharge;feature selection;random forest;HV cables;pattern recognition;
0 引言
高压电缆局部放电(简称局放)[1]新特征构建与优选,对局放和干扰的识别以及不同类型局放信号识别具有重要的意义。第1,对局放和干扰的识别,若干有效新特征的优选,能提升识别的精度,增强信号可视化效果。例如文献[2]中的T-W Mapping方法就是采用等效时间长度(equivalent time length,T)和等效带宽(equivalent bandwidth,W)2个核心特征来表征快脉冲和慢脉冲,以实现局放和干扰的识别。该研究成果被局放检测领域学者跟踪研究,并在工业监测系统中获得广泛的应用[3-6]。第2,对不同类型局放的识别,特征寻优能在保证识别率的前提下,提升识别的效率。现有文献中提出的局放特征包括:单个脉冲波形特征、统计特征、谱图特征、分形特征、序列特征、数字图像特征等,高达数百种[7-10]。将这些特征全部作为人工智能方法的输入参数,将带来维数灾,对识别方法的训练带来很大的难度。通过特征寻优,获得特征有效性排序,并进一步研究特征个数和识别率之间的关系,得到识别率上限对应的特征个数,将有效提升模式识别方法的效率。
图 1
基于随机森林的局放特征寻优流程
Fig. 1
Flowchart of random forest based optimal feature selection for partial discharge pattern recognition
第1阶段为实验数据获取和特征提取。本文对11 kV乙丙橡胶(ethylene propylene rubber,EPR)电缆设置5种类型的人工缺陷。加压后利用高频电流互感器(high frequency current transformer,HFCT)获取原始数据,并基于该数据提取了3500个干扰脉冲和3500个局放脉冲。其中局放样本分为5类,每类局放样本数目均为700个。详细的实验方法见文献[19-20]。然后基于提取的局放脉冲和干扰脉冲进行1维特征提取和2维、3维特征构建。共提取1维特征34个,构建2维、3维特征分别为119个和1082个,每个脉冲特征共计1235个。
表A1
局放信号和干扰信号的特征重要性排序
Tab. A1
Sorted scores of feature importance for PD signals and interference signals
表A2
5种类型局放信号特征重要性排序
Tab. A2
Sorted scores of feature importance for five types of PD signals
参考文献
[1]
张强,李成榕.X射线激励下局部放电的研究进展[J].,2017,32(8):22-32.ZhangQiang,LiChengrong.Review on X-Ray induced partial discharge[J].,2017,32(8):22-32(in Chinese).
[2]
Montanari GC,CavalliniA,PulettiF.A new approach to partial discharge testing of HV cable systems[J].,2006,22(1):14-23.
[3]
ContinA,PastoreS.Classification and separation of partial discharge signals by means of their auto-correlation function evaluation[J].,2009,16(6):1609-1622.
[5]
Chan JC,MaH,Saha TK.Time-frequency sparsity map on automatic partial discharge sources separation for power transformer condition assessment[J].,2015,22(4):2271-2283.
[6]
RenM,DongM,RenZ,et al.Transient earth voltage measurement in PD detection of artificial defect models in SF6[J].,2012,40(8):2002-2008.
[7]
李刚,汪可,张书琦,等.变压器油纸绝缘沿面放电缺陷发展特征及阶段识别方法[J].,2018,42(10):3451-3458.LiGang,WangKe,ZhangShuqi,et al.Evolution characteristics and stage recognition method of surface discharge defects of oil-paper in-sulation in transformer[J].,2018,42(10):3451-3458(in Chinese).
[8]
杨帆,杨旗,程鹏,等.电缆接头内部气隙放电缺陷下的绝缘劣化程度表征方法[J].,2017,32(2):24-32.YangFan,YangQi,ChengPeng,et al.Study of cracking extent for gap discharge in insulating material of power cable joint[J].,2017,32(2):24-32(in Chinese).
[9]
FiruziK,VakilianM,Darabad VP,et al.A novel method for differentiating and clustering multiple partial discharge sources using S transform and bag of words feature[J].,2018,24(6):3694-3702.
[10]
杨丰源,许永鹏,钱勇,等.相关分析技术在直流电压下XLPE电缆局部放电信号特征提取与筛选中的应用[J].,2018,42(5):1653-1660.YangFengyuan,XuYongpeng,QianYong,et al.Application of correlation analysis techniques in feature extraction and selection for DC partial discharge signals of XLPE cables[J].,2018,42(5):1653-1660(in Chinese).
[11]
张稳,盛万兴,刘科研,等.计及天气因素相关性的配电网故障风险等级预测方法[J].,2018,42(8):2391-2398.ZhangWen,ShengWanxing,LiuKeyan,et al.A prediction method of fault risk level for distribution network considering correlation of weather factors[J].,2018,42(8):2391-2398(in Chinese).
[12]
叶圣永,王晓茹,刘志刚,等.基于支持向量机的暂态稳定评估双阶段特征选择[J].,2010,30(31):28-34.YeShengyong,WangXiaoru,LiuZhigang,et al.Dual-stage feature selection for transient stability assessment based on support vector machine[J].,2010,30(31):28-34(in Chinese).
[13]
PedergnanaM,Marpu PR,Mura MD,et al.A novel technique for optimal feature selection in attribute profiles based on genetic algorithms[J].,2013,51(6):3514-3528.
[14]
PangH,George SL,HuiK,et al.Gene selection using iterative feature elimination random forests for survival outcomes[J].,2012,9(5):1422-1431.
[15]
Hasan M AM,NasserM,AhmadS,et al.Feature Selection for Intrusion Detection Using Random Forest[J].,2016,7(3):129-140.
[16]
Menze BH,Kelm BM,MasuchR,et al.A comparison of random forest and its Gini importance with standard chemometric methods for the feature selection and classification of spectral data[J].,2009,10(1):1-16.
[17]
BreimanL.Random forests[J].,2001,45(1):5-32.
[18]
方匡南,吴见彬,朱建平,等.随机森林方法研究综述[J].,2011,26(3):32-38.FangKuangnan,WuJianbin,ZhuJianping,et al.A review of technologies on random forests[J].,2011,26(3):32-38(in Chinese).
[19]
PengX,WenJ,LiZ,et al.SDMF based interference rejection and PD interpretation for simulated defects in HV cable diagnostics[J].,2017,24(1):83-91.
[20]
PengX,WenJ,LiZ,et al.Rough set theory applied to pattern recognition of partial discharge in noise affected cable data[J].,2017,24(1): 147-156.
[21]
张冠军,朱明晓,王彦博,等.基于可移动特高频天线阵列的变电站站域放电源检测与定位研究[J].,2017,37(10):2761-2773.ZhangGuanjun,ZhuMingxiao,WangYanbo,et al.Detection and localization of discharge sources in substation area using movable UHF antenna array[J].,2017,37(10):2761-2773(in Chinese).
[22]
周凯,赵世林,何珉,等.考虑短距离电缆中行波特性的振荡波局部放电定位方法[J].,2017,41(6):2047-2054.ZhouKai,ZhaoShilin,HeMin,et al.An oscillating wave test method based on traveling wave characteristics of partial discharges for defect location in short cables[J].,2017,41(6):2047-2054(in Chinese).