绝对值激活深度神经网络的串联故障电弧检测
|
摘要
串联故障电弧具有隐蔽性和随机性,发生时线路电流波形受负载类型的影响而具有复杂性,检测难度大,严重威胁用电系统安全。鉴于电流数据具有大量负值的特点,提出用绝对值函数作为激活函数改进AlexNet深度学习网络检测串联故障电弧,并分析了激活函数特性对串联故障电弧检测效果的影响。把实验采集的三类负载分别在正常和发生串联故障电弧状态下的共7 200组电流数据制作成训练集和测试集,并分别对使用四种激活函数的AlexNet网络进行训练和测试。实验结果显示,ELU激活的网络最高检测正确率为95.5%;而绝对值激活的网络效果最好,其平均检测正确率最高为97.25%,最低为93%,比ReLU激活的AlexNet网络最高88.75%的平均准确率高出最少4.25个百分点;而使用Sigmoid函数的网络不收敛。分析结果表明线性的激活数据特征有助于提高网络的检测准确率。
Series fault arc is difficult to detect due to its concealment random and complexity, which threatens the safety of domestic power supply system. Considering that current data have lots of negative values, an improved AlexNet deep learning network with absolute value function as activation function was proposed to detect series fault arc and the influence of activation function characteristics on the detection effect of series fault arc was analyzed. Totally 7 200 groups of current data of three kinds of load under normal and fault conditions were used as training and testing sets to train and test four AlexNets with four kinds of activation function respectively. Experimental results show that the maximum detection accuracy of ELU(Exponential Linear Unit)-activated network is 95.5%, while the performance of ABS(ABSolute value function)-activated network is the best, with the maximum average detection accuracy of 97.25% and the minimum detection accuracy of 93%, which is higher at least 4.25 percentage points than 88.75%, the maximum average detection accuracy of ReLU(Recified Linear Unit)-activated AlexNet, and Sigmoid-activated AlexNet does not converge. It is concluded that linear activation data characteristics improve the detection accuracy of network.
Series fault arc is difficult to detect due to its concealment random and complexity, which threatens the safety of domestic power supply system. Considering that current data have lots of negative values, an improved AlexNet deep learning network with absolute value function as activation function was proposed to detect series fault arc and the influence of activation function characteristics on the detection effect of series fault arc was analyzed. Totally 7 200 groups of current data of three kinds of load under normal and fault conditions were used as training and testing sets to train and test four AlexNets with four kinds of activation function respectively. Experimental results show that the maximum detection accuracy of ELU(Exponential Linear Unit)-activated network is 95.5%, while the performance of ABS(ABSolute value function)-activated network is the best, with the maximum average detection accuracy of 97.25% and the minimum detection accuracy of 93%, which is higher at least 4.25 percentage points than 88.75%, the maximum average detection accuracy of ReLU(Recified Linear Unit)-activated AlexNet, and Sigmoid-activated AlexNet does not converge. It is concluded that linear activation data characteristics improve the detection accuracy of network.
引文
[1] 中国公安部消防局.火灾统计[EB/OL].[2018- 09- 23] http://www.119.gov.cn/xiaofang/hztj/index.htm.
[2] 余琼芳.基于小波分析及数据融合的电气火灾预报系统及应用研究[D].秦皇岛:燕山大学,2013.
[3] 杨艺,董爱华,付永丽.低压故障电弧检测概述[J].低压电器,2009(5):1-4,27.
[4] YU Q F,ZHENG D Z,YANG Y,et al.Experimental study of on-line arc fault detection system based on LabWindows/CVI [J].Journal of Theoretical and Applied Information Technology,2012,42(1):150-155.
[5] 杨欣怡,蔡志远,程屾.低压故障电弧特征的研究与分析[C]// 第十四届沈阳科学学术年会论文集(理工农医).沈阳:沈阳市科学技术协会,2017:150-155.
[6] ZHAO Y Q,ZHANG X B,DONG Y J,et al.Characteristics analysis and detection of AC arc fault in SSPC based on wavelet transform[C]// Proceedings of the 2016 IEEE/CSAA International Conference on Aircraft Utility Systems.Beijing:IEEE/CSAA,2016:476-481.
[7] LI W L,HE K,LIU W,et al.A fast arc fault detection method for AC solid state power controllers in MEA [J].Chinese Journal of Aeronautics,2018,31(5):1119-1129.
[8] 李岚松,周越,熊翔,等.基于LS-SVM的航空故障电弧诊断[J].电器与能效管理技术,2018(10):45-49,59.
[9] 崔芮华,胡文达,耿丽恺.基于小波重构信号奇异点的航空故障电弧检测[J].电气传动,2018,48(6):69-72.
[10] 郭凤仪,高洪鑫,王智勇,等.基于ST-SVD-PCA的串联故障电弧特征提取方法[J].煤炭学报,2018,43(3):888-896.
[11] QI P,JOVANOVIC S,LEZAMA J,et al.Discrete wavelet transform optimal parameters estimation for arc fault detection in low-voltage residential power networks [J].Electric Power Systems Research,2017,143:130-139.
[12] CHAE S Y,PARK J J,OH S.Series DC arc fault detection algo-rithm for DC microgrids using relative magnitude comparison [J].IEEE Journal of Emerging and Selected Topics in Power Electronics,2016,4(4):1270-1278.
[13] LIU Y-W,WU C-J,WANG Y-C.Detection of serial arc fault on low-voltage indoor power lines by using radial basis function neural network [J].Electrical Power and Energy Systems,2016,83:149-157.
[14] 孟丹.基于深度学习的图像分类方法研究[D].上海:华东师范大学,2017.
[15] 胡聪,屈瑾瑾,许川佩,等.基于自适应池化的神经网络的服装图像识别[J].计算机应用,2018,38(8):2211-2217.
[16] 吕鸿蒙,赵地,迟学斌.基于增强AlexNet的深度学习的阿尔茨海默病的早期诊断[J].计算机科学,2017,44(Z6):50-60.
[17] PENG G,WANG Z,WEI Z,et al.Dimensionality reduction in deep learning for chest X-ray analysis of lung cancer[C]// ICACI 2018:Proceedings of the 2018 Tenth International Conference on Advanced Computational Intelligence.Piscataway:IEEE,2018:878-883.
[18] 陈东敏,姚剑敏.基于优化的并行AlexNet人脸特征点检测算法[J].信息技术与网络安全,2018,37(4):65-70.
[19] KARADUMAN G,KARAKOSE M,AKIN E.Deep learning based arc detection in pantograph-catenary systems [C] // ELECO 2017:Proceedings of the 2017 10th International Conference on Electrical and Electronics Engineering.Piscataway:IEEE,2017:904 -908.
[20] GUO M-F,ZENG X-D,CHEN D-Y,et al.Deep-learning-based earth fault detection using continuous wavelet transform and convolutional neural network in resonant grounding distribution systems [J].IEEE Sensors Journal,2018,18(3):1291-1300.
[21] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks [C]// Proceedings of the 25th International Conference on Neural Information Processing Systems.Nice,France:Curran Associates Inc,2012,1:1097-1105.
[22] CLEVERT D A,UNTERTHINER T,HOCHREITER S.Fast and accurate deep network learning by exponential linear units [C]// ICLR 2016:Proceedings of the 2016 International Conference on Learning Representations.San Juan,Puerto Rico:[s.n.],2016:1-14.
[23] 公安部沈阳消防研究所.GB 14287.4—2014 电气火灾监控系统第4部分:故障电弧探测器[S].北京:中华人民共和国国家质量监督检验检疫总局,中国国家标准化管理委员会,2015:13.
[2] 余琼芳.基于小波分析及数据融合的电气火灾预报系统及应用研究[D].秦皇岛:燕山大学,2013.
[3] 杨艺,董爱华,付永丽.低压故障电弧检测概述[J].低压电器,2009(5):1-4,27.
[4] YU Q F,ZHENG D Z,YANG Y,et al.Experimental study of on-line arc fault detection system based on LabWindows/CVI [J].Journal of Theoretical and Applied Information Technology,2012,42(1):150-155.
[5] 杨欣怡,蔡志远,程屾.低压故障电弧特征的研究与分析[C]// 第十四届沈阳科学学术年会论文集(理工农医).沈阳:沈阳市科学技术协会,2017:150-155.
[6] ZHAO Y Q,ZHANG X B,DONG Y J,et al.Characteristics analysis and detection of AC arc fault in SSPC based on wavelet transform[C]// Proceedings of the 2016 IEEE/CSAA International Conference on Aircraft Utility Systems.Beijing:IEEE/CSAA,2016:476-481.
[7] LI W L,HE K,LIU W,et al.A fast arc fault detection method for AC solid state power controllers in MEA [J].Chinese Journal of Aeronautics,2018,31(5):1119-1129.
[8] 李岚松,周越,熊翔,等.基于LS-SVM的航空故障电弧诊断[J].电器与能效管理技术,2018(10):45-49,59.
[9] 崔芮华,胡文达,耿丽恺.基于小波重构信号奇异点的航空故障电弧检测[J].电气传动,2018,48(6):69-72.
[10] 郭凤仪,高洪鑫,王智勇,等.基于ST-SVD-PCA的串联故障电弧特征提取方法[J].煤炭学报,2018,43(3):888-896.
[11] QI P,JOVANOVIC S,LEZAMA J,et al.Discrete wavelet transform optimal parameters estimation for arc fault detection in low-voltage residential power networks [J].Electric Power Systems Research,2017,143:130-139.
[12] CHAE S Y,PARK J J,OH S.Series DC arc fault detection algo-rithm for DC microgrids using relative magnitude comparison [J].IEEE Journal of Emerging and Selected Topics in Power Electronics,2016,4(4):1270-1278.
[13] LIU Y-W,WU C-J,WANG Y-C.Detection of serial arc fault on low-voltage indoor power lines by using radial basis function neural network [J].Electrical Power and Energy Systems,2016,83:149-157.
[14] 孟丹.基于深度学习的图像分类方法研究[D].上海:华东师范大学,2017.
[15] 胡聪,屈瑾瑾,许川佩,等.基于自适应池化的神经网络的服装图像识别[J].计算机应用,2018,38(8):2211-2217.
[16] 吕鸿蒙,赵地,迟学斌.基于增强AlexNet的深度学习的阿尔茨海默病的早期诊断[J].计算机科学,2017,44(Z6):50-60.
[17] PENG G,WANG Z,WEI Z,et al.Dimensionality reduction in deep learning for chest X-ray analysis of lung cancer[C]// ICACI 2018:Proceedings of the 2018 Tenth International Conference on Advanced Computational Intelligence.Piscataway:IEEE,2018:878-883.
[18] 陈东敏,姚剑敏.基于优化的并行AlexNet人脸特征点检测算法[J].信息技术与网络安全,2018,37(4):65-70.
[19] KARADUMAN G,KARAKOSE M,AKIN E.Deep learning based arc detection in pantograph-catenary systems [C] // ELECO 2017:Proceedings of the 2017 10th International Conference on Electrical and Electronics Engineering.Piscataway:IEEE,2017:904 -908.
[20] GUO M-F,ZENG X-D,CHEN D-Y,et al.Deep-learning-based earth fault detection using continuous wavelet transform and convolutional neural network in resonant grounding distribution systems [J].IEEE Sensors Journal,2018,18(3):1291-1300.
[21] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks [C]// Proceedings of the 25th International Conference on Neural Information Processing Systems.Nice,France:Curran Associates Inc,2012,1:1097-1105.
[22] CLEVERT D A,UNTERTHINER T,HOCHREITER S.Fast and accurate deep network learning by exponential linear units [C]// ICLR 2016:Proceedings of the 2016 International Conference on Learning Representations.San Juan,Puerto Rico:[s.n.],2016:1-14.
[23] 公安部沈阳消防研究所.GB 14287.4—2014 电气火灾监控系统第4部分:故障电弧探测器[S].北京:中华人民共和国国家质量监督检验检疫总局,中国国家标准化管理委员会,2015:13.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |