林烨敏,王宁,邱荣杰,汤宇超,周冠群,李泽洲,王中亚
LIN Yemin,WANG Ning,QIU Rongjie,TANG Yuchao,ZHOU Guanqun,LI Zezhou,WANG Zhongya

• 1:国网浙江省电力有限公司紧水滩水力发电厂
• 2:宁波工业互联网研究院

摘要(Abstract)：

水电站机组日常检修、维护及异常检测的工作量巨大，传统人工监测的工作方式容易导致异常问题被遗漏或误判，采用深度学习算法对数据建模并监测异常情况可以降低成本，提升安全可靠性。结合Transformer网络对长期序列高效准确建模的能力以及GAN（生成对抗网络）架构的数据生成训练策略，利用TransGAN模型对水电机组监测数据进行生成式建模，并主动发现异常数据点。TransGAN模型在水电站机组实测中达到了97.76%的查准率和99.23%的查全率，异常点检出延迟低于0.1 s，实现了实时高精度异常监控功能。
The workloads of routine repair, maintenance, and abnormality detection of hydropower units are heavy.Therefore, traditional manual monitoring may leave out or misjudge abnormalities. Deep learning algorithms are used for data modeling and monitoring abnormalities to reduce costs and improve safety and reliability. With the help of the Transformer neural networks, the efficient and accurate modeling capacity of long-term sequences and the GAN(generative adversarial network) architecture data are used to generate a training strategy. A TransGAN model is used for generative modeling of the measured data of hydropower units and proactively detects abnormal data points. The TransGAN model achieves a detection accuracy rate of 97.76% and a recall rate of 99.23% in hydropower data measurement. The anomaly detection delay is less than 0.1 s. The real-time high-precision anomaly monitoring function is realized.

关键词(KeyWords)： 水电机组;异常监测;数据降维;Transformer;生成对抗网络
hydropower units;anomaly detection;dimensionality reduction;Transformer;GAN

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司科技项目（B311JS210002）

作者(Author): 林烨敏,王宁,邱荣杰,汤宇超,周冠群,李泽洲,王中亚
LIN Yemin,WANG Ning,QIU Rongjie,TANG Yuchao,ZHOU Guanqun,LI Zezhou,WANG Zhongya

DOI: 10.19585/j.zjdl.202301014

参考文献(References)：

• [1]李文清，周丰兴.水电机组故障诊断分析系统[J].浙江电力，2002,21(5):46-48.LI Wenqing,ZHOU Fengxing.Fault diagnosis and analysis system for hydropower units[J]. Zhejiang Electric Power,2002,21(5):46-48.
• [2] CHANDOLA V,BANERJEE A,KUMAR V.Anomaly detection:a survey[J].ACM Computing Surveys,2009,41(3):1-58.
• [3] BREUNIG M,KRIEGEL H,NG R,et al.LOF:identifying density-based local outliers.2000.
• [4] SCH?LKOPF B,PLATT J C,SHAWE-TAYLOR J C,et al.Estimating the support of a high-dimensional distribution[J].Neural Computation,2001,13(7):1443-1471.
• [5] SU Y,ZHAO Y J,NIU C H,et al.Robust anomaly detection for multivariate time series through stochastic recurrent neural network[C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery&Data Mining. Anchorage,AK,USA. New York:ACM,2019:2828-2837.
• [6] SHEN L F, LI Z C,KWOK J T.Timeseries anomaly detection using temporal hierarchical one-class network. In Hugo Larochelle,Marc’Aurelio Ranzato, Raia Hadsell,MariaFlorina Balcan,and Hsuan-Tien Lin(eds.),NeurIPS,2020.
• [7] VASWANI A,SHAZEER N M,PARMAR N,et al.Attention is all You need.2017.
• [8] HUANG S H,LIU Y,FUNG C,et al.HitAnomaly:hierarchical transformers for anomaly detection in system log[J].IEEE Transactions on Network and Service Management,2020,17(4):2064-2076.
• [9] CHEN Z K,CHEN D S,ZHANG X,et al.Learning graph structures with transformer for multivariate time-series anomaly detection in IoT[J]. IEEE Internet of Things Journal,2022,9(12):9179-9189.
• [10] GOODFELLOW Ian,POUGETABADIE Jean,MIRZA Mehdi,et al. Generative adversarial nets. In Advances in neural information processing systems,2014:2672-2680.
• [11] ZHOU B,LIU S H,HOOI B,et al.BeatGAN:anomalous rhythm detection using adversarially generated time series[C]//Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence. Macao,China.California:International Joint Conferences on Artificial Intelligence Organization,2019:4433-4439.
• [12]吴晓婷，闫德勤.数据降维方法分析与研究[J].计算机应用研究，2009,36(8):2832-2835.WU Xiaoting,YAN Deqin. Analysis and research on method of data dimensionality reduction[J]. Application Research of Computers,2009,26(8):2832-2835.
• [13]李航.统计学习方法[M].2版.北京：清华大学出版社，2019.
• [14] LIU F,DONG T,HOU T,et al.A hybrid short-term load forecasting model based on improved fuzzy C-means clustering,random forest and deep neural networks[J].IEEE Access,2021,9:59754-59765.
• [15]谭振宁.基于深度学习的时序预测和分类[D].广州：华南理工大学，2020.TAN Zhenning. Time series prediction and classification based on deep learning[D].Guangzhou:South China University of Technology,2020.