作者：牛罡1, 张凯2, 郭祥富2, 胡军星1, 袁林峰3, 徐勇明3, 史建勋3, 张志友4
作者单位：1. 河南九域腾龙信息工程有限公司，河南 郑州 450052;
2. 国网河南省电力公司，河南 郑州 450000;
3. 国网浙江嘉善县供电有限公司，浙江 嘉善 314100;
4. 东南大学电气工程学院，江苏 南京 210096
关键词：格拉米角场;图卷积神经网络;电力系统;电能质量
摘要：
可再生能源并网引起的电能质量问题往往伴随着复杂的电能质量扰动，因此电能质量的扰动类型识别是后续电能污染控制的首要任务。该文针对复杂电能质量扰动的分类问题，提出一种基于格拉米角场和图卷积神经网络的复杂电能质量扰动识别方法。首先，通过搭建直驱型风力发电动态模拟实验平台模拟生成不同的单一扰动和复杂扰动信号，将原始一维时间序列的电能质量扰动数据通过格拉米角场值的密度分布进行表示，生成二维格拉米角场值密度图；然后，通过图卷积神经网络对输入图像进行无监督特征学习，自动提取数据特征的稀疏特征表达；最后，通过softmax分类器进行微调训练，并输出电能质量扰动事件分类结果。通过仿真和硬件实验表明：对于复杂电能质量的扰动类型识别，该方法能保持94.87%左右的识别准确率，在算法训练速度和识别精度方面，优于其他常见深度学习算法，具有良好的工程应用前景。
Complex power quality disturbance identification based on Gramian angular field and graph convolution neural network
NIU Gang1, ZHANG Kai2, GUO Xiangfu2, HU Junxing1, YUAN Linfeng3, XU Yongming3, SHI Jianxun3, ZHANG Zhiyou4
1. Henan Jiuyu Tenglong Information Engineering Co., Ltd., Zhengzhou 450052, China;
2. State Grid Hennan Electric Power Company, Zhengzhou 450000, China;
3. State Grid Jiashan Power Supply Company, Jiashan 314100, China;
4. School of Electrical Engineering, Southeast University, Nanjing 210096, China
Abstract: Power quality problems caused by renewable energy integration are often accompanied by complex power quality disturbances, so the identification of power quality disturbance types is the primary task of power pollution control. Aiming at the classification of complex power quality disturbances, this paper proposes a recognition method of complex power quality disturbances based on Gramian angular field and graph convolution neural network. Firstly, different single disturbance and complex disturbance signals are generated by building a direct drive wind power generation dynamic simulation experimental platform, and the power quality disturbance data of the original one-dimensional time series is represented by the density distribution of the Gramian angular field value to generate a two-dimensional Gramian angular field value density map; then, the unsupervised feature analysis of the input image is performed by graph convolution neural network Finally, the softmax classifier is used for fine-tuning training, and the classification results of power quality disturbance events are output. The simulation and hardware experiments show that this method can keep the recognition accuracy of 94.87% for the disturbance type recognition of complex power quality. In terms of algorithm training speed and recognition accuracy, there are other common deep learning algorithms, which has a good engineering application prospect.
Keywords: Gramian angular field;graph convolution neural network;power system;power qualitys
2022, 48(7):169-176 收稿日期: 2021-03-31;收到修改稿日期: 2021-06-01
基金项目:
作者简介: 牛罡（1975-），男，河南登封市人，高级工程师，研究方向为电气系统与自动化
参考文献
[1] 李建林, 袁晓冬, 郁正纲, 等. 利用储能系统提升电网电能质量研究综述[J]. 电力系统自动化, 2019, 43(8): 15-24
[2] 杨新法, 苏剑, 吕志鹏, 等. 微电网技术综述[J]. 中国电机工程学报, 2014, 34(1): 57-70
[3] 何吉彪, 程浩忠. 含微网配电网规划中的电能质量综合评估[J]. 电网技术, 2012, 36(8): 209-214
[4] 徐艳春, 高永康, 李振兴, 等. 基于VMD初始化S变换的混合动力系统电能质量扰动检测与分类[J]. 中国电机工程学报, 2019, 39(16): 4786-4798+4980
[5] 徐艳春, 高永康, 李振兴, 等. 改进LMD算法在微电网电能质量扰动信号检测中的应用[J]. 电网技术, 2019, 43(1): 332-341
[6] 王仁明, 汪宏阳, 张赟宁, 等. 基于分段改进S变换和随机森林的复合电能质量扰动识别方法[J]. 电力系统保护与控制, 2020, 48(7): 19-28
[7] 黄建明, 瞿合祚, 李晓明. 基于短时傅里叶变换及其谱峭度的电能质量混合扰动分类[J]. 电网技术, 2016, 40(10): 3184-3191
[8] LU C G, HUO Z Q, CHENG X, et al. Distributed optical fiber sensing intrusion pattern recognition based on GAF and CNN[J]. Journal of Lightwave Technology, 2020, 38(15): 4174-4182
[9] HU C, CHEN Z Y, DONG X C, et al. Multistatic geosynchronous SAR resolution analysis and grating lobe suppression based on array spatial ambiguity function[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(9): 6020-6038
[10] LI C, QIN X, XU X, et al. Scalable graph convolutional networks with fast localized spectral filter for directed graphs[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 38(19): 105634-105644
[11] 徐冰冰, 岑科廷, 黄俊杰, 等. 图卷积神经网络综述[J]. 计算机学报, 2020, 43(5): 755-780
[12] HU Z X, WANG Y, GE M F, et al. Data-driven fault diagnosis method based on compressed sensing and improved multiscale network[J]. IEEE Transactions on Industrial Electronics, 2020, 67(10): 3216-3225
[13] MA S L, CHEN M X, WU J W, et al. High-voltage circuit breaker fault diagnosis using a hybrid feature transformation approach based on random forest and stacked autoencoder[J]. IEEE Transactions on Industrial Electronics, 2019, 66(15): 9777-9788
[14] 武昭旭, 杨岸, 祝龙记. 基于循环神经网络的电能质量扰动识别[J]. 电力系统保护与控制, 2020, 48(18): 88-94
[15] 马杰, 孙宁, 任志强, 等. 基于多特征组合的电能质量扰动识别方法[J]. 电力电容器与无功补偿, 2019, 40(2): 125-130
[16] 张殷, 武利会, 范心明, 等. 基于改进S变换和GA-SVM的电能质量扰动识别与分类[J]. 广东电力, 2021, 34(5): 99-106
