采用改进生成式对抗网络的电力系统量测缺失数据重建方法
|
摘要
量测数据的采集、传输、转换各个环节都有可能发生故障或受到干扰,导致数据出现缺失。传统重建方法仅考虑单一数据分布规律,忽略了电力系统中量测点、采集变量之间的相关性及历史的负荷变化规律,重建精度低。该文提出了基于改进生成式对抗网络(wassersteingenerative adversarial networks,WGAN)的量测缺失值重建方法,并设计了适用于该问题的WGAN网络结构。通过WGAN的无监督训练,神经网络将自动学习到量测之间相关性、负荷波动规律等难以显式建模的复杂时空关系。利用真实性约束及上下文相似性约束优化隐变量,使得训练后的生成器将能够生成高精度的重建数据。文中方法完全依靠数据驱动,不涉及显式建模步骤,在大量量测出现缺失的情况下仍具有较高的重建精度。算例中分析了量测缺失数量与重建误差之间的关系,证明了文中方法性能稳定。对于算例中长期缺失的特定量测,文中方法所重建的数据能够体现量测真实的时序特性。
The collection, transmission and conversion of measured data may be interrupted or disturbed, leading to the loss of data. The traditional reconstruction method only considers the distribution of single data, ignores the correlation between measurement points, measured variables and the historical load change in power system. In this paper, an improved Wasserstein generating adversarial networks(WGAN) method for the reconstruction of missing value was proposed, and a WGAN network structure was designed. Through the unsupervised training of WGAN, the neural network could automatically learn the complex space-time relation which was difficult to model explicitly, such as the correlation between measurement and load fluctuation rule. The hidden variables were optimized by using the real constraint and context similarity constraint, so that the generator after training could generate high precision reconstruction data. The method in this paper is completely data-driven and does not involve explicit modeling steps. It still has high reconstruction accuracy when a large number of measurements are missing. In the calculation example, the relationship between the number of measurements missing and the reconstruction error is analyzed in the example, and it's proved that the data reconstructed can reflect the real time sequence characteristics of the measurement for the long term missing specific measurement.
The collection, transmission and conversion of measured data may be interrupted or disturbed, leading to the loss of data. The traditional reconstruction method only considers the distribution of single data, ignores the correlation between measurement points, measured variables and the historical load change in power system. In this paper, an improved Wasserstein generating adversarial networks(WGAN) method for the reconstruction of missing value was proposed, and a WGAN network structure was designed. Through the unsupervised training of WGAN, the neural network could automatically learn the complex space-time relation which was difficult to model explicitly, such as the correlation between measurement and load fluctuation rule. The hidden variables were optimized by using the real constraint and context similarity constraint, so that the generator after training could generate high precision reconstruction data. The method in this paper is completely data-driven and does not involve explicit modeling steps. It still has high reconstruction accuracy when a large number of measurements are missing. In the calculation example, the relationship between the number of measurements missing and the reconstruction error is analyzed in the example, and it's proved that the data reconstructed can reflect the real time sequence characteristics of the measurement for the long term missing specific measurement.
引文
[1]丁宏恩,高宗和,苏大威,等.混合量测状态估计相角参考点坏数据问题的处理方法[J].电力系统自动化,2014,38(9):132-136.Ding Hongen,Gao Zonghe,Su Dawei,et al.Solution to bad data problem of phase angle reference bus for state estimation with hybrid measurement[J].Automation of Electric Power Systems,2014,38(9):132-136(in Chinese).
[2]严英杰,盛戈皞,陈玉峰,等.基于时间序列分析的输变电设备状态大数据清洗方法[J].电力系统自动化,2015,39(7):138-144.Yan Yingjie,Sheng Gehao,Chen Yufeng,et al.Cleaning method for big data of power transmission and transformation equipment state based on time sequence analysis[J].Automation of Electric Power Systems,2015,39(7):138-144(in Chinese).
[3]王德文,孙志伟.电力用户侧大数据分析与并行负荷预测[J].中国电机工程学报,2015,35(3):527-537.Wang Dewen,Sun Zhiwei.Big data analysis and parallel load forecasting of electric power user side[J].Proceedings of the CSEE,2015,35(3):527-537(in Chinese).
[4]张素香,赵丙镇,王风雨,等.海量数据下的电力负荷短期预测[J].中国电机工程学报,2015,35(1):37-42.Zhang Suxiang,Zhao Bingzhen,Wang Fengyu,et al.Short-term power load forecasting based on big data[J].Proceedings of the CSEE,2015,35(1):37-42(in Chinese).
[5]张东英,李伟花,刘燕华,等.风电场有功功率异常运行数据重构方法[J].电力系统自动化,2014,38(5):14-18,24.Zhang Dongying,Li Weihua,Liu Yanhua,et al.Reconstruction method of active power historical operating data for wind farm[J].Automation of Electric Power Systems,2014,38(5):14-18,24(in Chinese).
[6]龙婧,刘伟,殷胜.基于机器学习的电网设备档案数据异常诊断研究[J].电力信息与通信技术,2018,16(7):21-27.Long Jing,Liu Wei,Yin Sheng.Research on abnormal diagnosis for power grid equipment archival data based on machine learning[J].Electric Power Information and Communication Technology,2018,16(7):21-27(in Chinese).
[7]Miranda V,Krstulovic J,Keko H,et al.Reconstructing missing data in state estimation with autoencoders[J].IEEE Transactions on Power Systems,2012,27(2):604-611.
[8]Azarkhail M,Woytowitz P.Uncertainty management in model-based imputation for missing data[C]//Proceedings of 2013 Proceedings Annual Reliability and Maintainability Symposium.Orlando:IEEE,2013:1-7.
[9]王坤峰,苟超,段艳杰,等.生成式对抗网络GAN的研究进展与展望[J].自动化学报,2017,43(3):321-332.Wang Kunfeng,Gou Chao,Duan Yanjie,et al.Generative adversarial networks:The state of the art and beyond[J].Acta Automatica Sinica,2017,43(3):321-332(in Chinese).
[10]姚乃明,郭清沛,乔逢春,等.基于生成式对抗网络的鲁棒人脸表情识别[J].自动化学报,2018,44(5):865-877.Yao Naiming,Guo Qingpei,Qiao Fengchun,et al.Robust facial expression recognition with generative adversarial networks[J].Acta Automatica Sinica,2018,44(5):865-877(in Chinese).
[11]Goodfellow I J,Pouget-Abadie J,Mirza M,et al.Generative adversarial nets[C]//Proceedings of the27thInternational Conference on Neural Information Processing Systems.Montreal:MIT Press,2014:2672-2680.
[12]Goodfellow I.NIPS 2016 tutorial:Generative adversarial networks[J].arXiv preprint arXiv:1701.00160,2016.
[13]唐贤伦,杜一铭,刘雨微,等.基于条件深度卷积生成对抗网络的图像识别方法[J].自动化学报,2018,44(5):855-864.Tang Xianlun,Du Yiming,Liu Yuwei,et al.Image recognition with conditional deep convolutional generative adversarial networks[J].Acta Automatica Sinica,2018,44(5):855-864(in Chinese).
[14]刘一敏,蒋建国,齐美彬,等.合生成对抗网络和姿态估计的视频行人再识别方法[J/OL].自动化学报:1-9[2018-12-19].https://doi.org/10.16383/j.aas.c180054.Liu Yimin,Jiang Jianguo,Qi Meibin,et al.Video-based person re-identication method based on GAN and pose estimation[J/OL].Acta Automatica Sinica:1-9[2018-12-19].https://doi.org/10.16383/j.aas.c180054(in Chinese).
[15]Ledig C,Theis L,Huszár F,et al.Photo-realistic single image super-resolution using a generative adversarial network[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition.Honolulu:IEEE,2017:105-114.
[16]Yeh R A,Chen C,Lim T Y,et al.Semantic image inpainting with deep generative models[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition.Honolulu:IEEE,2017.
[17]Arjovsky M,Chintala S,Bottou L.Wasserstein GAN[J].arXiv preprint arXiv:1701.07875,2017.
[18]Ioffe S,Szegedy C.Batch normalization:accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on Machine Learning.Lille,France:PMLR,2015,37:448-456.
[19]Xu Bing,Wang Naiyan,Chen Tianqi,et al.Empirical evaluation of rectified activations in convolutional network[C]//ICML Deep Learning Workshop.Lille Grand Palais:ICML,2015.
[20]Kingma D P,Ba J.Adam:A method for stochastic optimization[J].arXiv preprint arXiv:1412.6980,2014:434-435.
[21]Pathak D,Kr?henbühl P,Donahue J,et al.Context encoders:feature learning by inpainting[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas:IEEE,2016:2536-2544.
[22]Dam Q B,Meliopoulos A P S,Heydt G T,et al.Abreaker-oriented,three-phase IEEE 24-substation test system[J].IEEE Transactions on Power Systems,2010,25(1):59-67.
[2]严英杰,盛戈皞,陈玉峰,等.基于时间序列分析的输变电设备状态大数据清洗方法[J].电力系统自动化,2015,39(7):138-144.Yan Yingjie,Sheng Gehao,Chen Yufeng,et al.Cleaning method for big data of power transmission and transformation equipment state based on time sequence analysis[J].Automation of Electric Power Systems,2015,39(7):138-144(in Chinese).
[3]王德文,孙志伟.电力用户侧大数据分析与并行负荷预测[J].中国电机工程学报,2015,35(3):527-537.Wang Dewen,Sun Zhiwei.Big data analysis and parallel load forecasting of electric power user side[J].Proceedings of the CSEE,2015,35(3):527-537(in Chinese).
[4]张素香,赵丙镇,王风雨,等.海量数据下的电力负荷短期预测[J].中国电机工程学报,2015,35(1):37-42.Zhang Suxiang,Zhao Bingzhen,Wang Fengyu,et al.Short-term power load forecasting based on big data[J].Proceedings of the CSEE,2015,35(1):37-42(in Chinese).
[5]张东英,李伟花,刘燕华,等.风电场有功功率异常运行数据重构方法[J].电力系统自动化,2014,38(5):14-18,24.Zhang Dongying,Li Weihua,Liu Yanhua,et al.Reconstruction method of active power historical operating data for wind farm[J].Automation of Electric Power Systems,2014,38(5):14-18,24(in Chinese).
[6]龙婧,刘伟,殷胜.基于机器学习的电网设备档案数据异常诊断研究[J].电力信息与通信技术,2018,16(7):21-27.Long Jing,Liu Wei,Yin Sheng.Research on abnormal diagnosis for power grid equipment archival data based on machine learning[J].Electric Power Information and Communication Technology,2018,16(7):21-27(in Chinese).
[7]Miranda V,Krstulovic J,Keko H,et al.Reconstructing missing data in state estimation with autoencoders[J].IEEE Transactions on Power Systems,2012,27(2):604-611.
[8]Azarkhail M,Woytowitz P.Uncertainty management in model-based imputation for missing data[C]//Proceedings of 2013 Proceedings Annual Reliability and Maintainability Symposium.Orlando:IEEE,2013:1-7.
[9]王坤峰,苟超,段艳杰,等.生成式对抗网络GAN的研究进展与展望[J].自动化学报,2017,43(3):321-332.Wang Kunfeng,Gou Chao,Duan Yanjie,et al.Generative adversarial networks:The state of the art and beyond[J].Acta Automatica Sinica,2017,43(3):321-332(in Chinese).
[10]姚乃明,郭清沛,乔逢春,等.基于生成式对抗网络的鲁棒人脸表情识别[J].自动化学报,2018,44(5):865-877.Yao Naiming,Guo Qingpei,Qiao Fengchun,et al.Robust facial expression recognition with generative adversarial networks[J].Acta Automatica Sinica,2018,44(5):865-877(in Chinese).
[11]Goodfellow I J,Pouget-Abadie J,Mirza M,et al.Generative adversarial nets[C]//Proceedings of the27thInternational Conference on Neural Information Processing Systems.Montreal:MIT Press,2014:2672-2680.
[12]Goodfellow I.NIPS 2016 tutorial:Generative adversarial networks[J].arXiv preprint arXiv:1701.00160,2016.
[13]唐贤伦,杜一铭,刘雨微,等.基于条件深度卷积生成对抗网络的图像识别方法[J].自动化学报,2018,44(5):855-864.Tang Xianlun,Du Yiming,Liu Yuwei,et al.Image recognition with conditional deep convolutional generative adversarial networks[J].Acta Automatica Sinica,2018,44(5):855-864(in Chinese).
[14]刘一敏,蒋建国,齐美彬,等.合生成对抗网络和姿态估计的视频行人再识别方法[J/OL].自动化学报:1-9[2018-12-19].https://doi.org/10.16383/j.aas.c180054.Liu Yimin,Jiang Jianguo,Qi Meibin,et al.Video-based person re-identication method based on GAN and pose estimation[J/OL].Acta Automatica Sinica:1-9[2018-12-19].https://doi.org/10.16383/j.aas.c180054(in Chinese).
[15]Ledig C,Theis L,Huszár F,et al.Photo-realistic single image super-resolution using a generative adversarial network[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition.Honolulu:IEEE,2017:105-114.
[16]Yeh R A,Chen C,Lim T Y,et al.Semantic image inpainting with deep generative models[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition.Honolulu:IEEE,2017.
[17]Arjovsky M,Chintala S,Bottou L.Wasserstein GAN[J].arXiv preprint arXiv:1701.07875,2017.
[18]Ioffe S,Szegedy C.Batch normalization:accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on Machine Learning.Lille,France:PMLR,2015,37:448-456.
[19]Xu Bing,Wang Naiyan,Chen Tianqi,et al.Empirical evaluation of rectified activations in convolutional network[C]//ICML Deep Learning Workshop.Lille Grand Palais:ICML,2015.
[20]Kingma D P,Ba J.Adam:A method for stochastic optimization[J].arXiv preprint arXiv:1412.6980,2014:434-435.
[21]Pathak D,Kr?henbühl P,Donahue J,et al.Context encoders:feature learning by inpainting[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas:IEEE,2016:2536-2544.
[22]Dam Q B,Meliopoulos A P S,Heydt G T,et al.Abreaker-oriented,three-phase IEEE 24-substation test system[J].IEEE Transactions on Power Systems,2010,25(1):59-67.