第16卷第6期 智能系统学报 Vol.16 No.6 2021年11月 CAAI Transactions on Intelligent Systems Now.2021 D0:10.11992tis.202009020 网络出版地址：https:/kns.cnki.net/kcms/detail/23.1538.TP.20210831.1251.004html 基于AC-GAN数据重构的风电机组 主轴承温度监测方法 尹诗2，侯国莲2，胡晓东2，周继威 (1.中能电力科技开发有限公司，北京100034：2.华北电力大学控制与计算机工程学院，北京102206) 摘要：为更好地识别风电机组主轴承运行状态，提出了一种基于辅助分类生成对抗网络(auxiliary classifier generative adversarial networks,AC-GAN)的数据重构算法对风电机组主轴承温度进行监测。首先，利用采集与 监视控制系统(supervisory control and data acquisition,SCADA)时序数据建立基于轻型梯度增强学习器((light gradient boosting machine,LightGBM)的主轴承温度预测模型，并计算其残差特征。其次，利用统计过程控制 (statistical process control,SPC)方法对主轴承温度异常残差在控制线范围内进行筛选，并利用AC-GAN算法对 残差进行重构。最后，分别提取主轴承温度正常和异常的残差特征，建立基于自然梯度提升(natural gradient boosting,NGBoost)的主轴承状态监测模型。实验结果表明，该方法对主轴承运行状态判断准确度高达87.5% 能够有效地监测风电机组轴承类运行状态。 关键词：轻型梯度增强学习器；辅助分类生成对抗网络；自然梯度提升；风电机组；主轴承；状态监测；数据重 构；温度残差 中图分类号：TP8,TK83文献标志码：A文章编号：1673-4785(2021)06-1106-11 中文引用格式：尹诗，侯国莲，胡晓东，等.基于AC-GAN数据重构的风电机组主轴承温度监测方法机.智能系统学报，2021， 16(6):1106-1116. 英文引用格式：YIN Shi,.HOU Guolian,HU Xiaodong,.etal.Temperature monitoring method of the main bearing of wind turbine based on AC-GAN data reconstruction[J).CAAI transactions on intelligent systems,2021,16(6):1106-1116. Temperature monitoring method of the main bearing of wind turbine based on AC-GAN data reconstruction YIN Shi,HOU Guolian',HU Xiaodong',ZHOU Jiwei' (1.ZhongNeng Power-Tech Development Co.,LTD,Beijing 100034,China;2.College of Control and Computer Engineering,North China Electric Power University,Beijing 102206,China) Abstract:To better identify the operating status of the main bearing of wind turbines,a data reconstruction algorithm based on auxiliary classifier generative adversarial networks(AC-GAN)is proposed to monitor the temperature of the main bearing of the wind turbine.First,this work uses the time series data of the supervisory control and data acquisi- tion system to establish the wind turbine's main bearing temperature prediction model based on the light gradient boost- ing machine and calculates its residual characteristics.Second,the statistical process control(SPC)method is used to screen abnormal temperature residuals of the main bearing within the control line,and the AC-GAN algorithm is used to reconstruct the residual sequence.Finally,normal and abnormal temperature residual characteristics of the main bearing are extracted,and the main bearing status monitoring model based on the natural gradient boosting algorithm is estab- lished.Experimental results show that the accuracy of the method for judging the operating state of the main bearing is as high as 87.5%,for which the algorithm can effectively monitor the running state of wind turbine bearings. Keywords:light gradient boosting machine;auxiliary classifier generative adversarial networks;natural gradient boost- ing:wind turbines:main bearing:condition monitoring:data reconstruction:temperature residual 由于风电机组所处运行环境恶劣，受气象、 收稿日期：2020-09-15.网络出版日期：2021-08-31. 基金项目：国家自然科学基金项目(61973116). 设备老化等多种不确定因素的影响，容易出现性 通信作者：尹诗.E-mail:yinshi502@163.com 能与运行状态劣化，从而造成关键部件失效。风DOI: 10.11992/tis.202009020 网络出版地址: https://kns.cnki.net/kcms/detail/23.1538.TP.20210831.1251.004.html 基于 AC-GAN 数据重构的风电机组 主轴承温度监测方法 尹诗1,2，侯国莲2 ，胡晓东2 ，周继威1 （1. 中能电力科技开发有限公司，北京 100034; 2. 华北电力大学 控制与计算机工程学院，北京 102206） 摘 要：为更好地识别风电机组主轴承运行状态，提出了一种基于辅助分类生成对抗网络 (auxiliary classifier generative adversarial networks, AC-GAN) 的数据重构算法对风电机组主轴承温度进行监测。首先，利用采集与 监视控制系统 (supervisory control and data acquisition, SCADA) 时序数据建立基于轻型梯度增强学习器 (light gradient boosting machine, LightGBM) 的主轴承温度预测模型，并计算其残差特征。其次，利用统计过程控制 (statistical process control, SPC) 方法对主轴承温度异常残差在控制线范围内进行筛选，并利用 AC-GAN 算法对 残差进行重构。最后，分别提取主轴承温度正常和异常的残差特征，建立基于自然梯度提升 (natural gradient boosting, NGBoost) 的主轴承状态监测模型。实验结果表明，该方法对主轴承运行状态判断准确度高达 87.5%， 能够有效地监测风电机组轴承类运行状态。 关键词：轻型梯度增强学习器；辅助分类生成对抗网络；自然梯度提升；风电机组；主轴承；状态监测；数据重 构；温度残差 中图分类号：TP8; TK83 文献标志码：A 文章编号：1673−4785(2021)06−1106−11 中文引用格式：尹诗, 侯国莲, 胡晓东, 等. 基于 AC-GAN 数据重构的风电机组主轴承温度监测方法 [J]. 智能系统学报, 2021, 16(6): 1106–1116. 英文引用格式：YIN Shi, HOU Guolian, HU Xiaodong, et al. Temperature monitoring method of the main bearing of wind turbine based on AC-GAN data reconstruction[J]. CAAI transactions on intelligent systems, 2021, 16(6): 1106–1116. Temperature monitoring method of the main bearing of wind turbine based on AC-GAN data reconstruction YIN Shi1,2 ，HOU Guolian2 ，HU Xiaodong2 ，ZHOU Jiwei1 (1. ZhongNeng Power-Tech Development Co., LTD, Beijing 100034, China; 2. College of Control and Computer Engineering, North China Electric Power University, Beijing 102206, China) Abstract: To better identify the operating status of the main bearing of wind turbines, a data reconstruction algorithm based on auxiliary classifier generative adversarial networks (AC-GAN) is proposed to monitor the temperature of the main bearing of the wind turbine. First, this work uses the time series data of the supervisory control and data acquisi￾tion system to establish the wind turbine’s main bearing temperature prediction model based on the light gradient boost￾ing machine and calculates its residual characteristics. Second, the statistical process control (SPC) method is used to screen abnormal temperature residuals of the main bearing within the control line, and the AC-GAN algorithm is used to reconstruct the residual sequence. Finally, normal and abnormal temperature residual characteristics of the main bearing are extracted, and the main bearing status monitoring model based on the natural gradient boosting algorithm is estab￾lished. Experimental results show that the accuracy of the method for judging the operating state of the main bearing is as high as 87.5%, for which the algorithm can effectively monitor the running state of wind turbine bearings. Keywords: light gradient boosting machine; auxiliary classifier generative adversarial networks; natural gradient boost￾ing; wind turbines; main bearing; condition monitoring; data reconstruction; temperature residual 由于风电机组所处运行环境恶劣，受气象、 设备老化等多种不确定因素的影响，容易出现性 能与运行状态劣化，从而造成关键部件失效。风 收稿日期：2020−09−15. 网络出版日期：2021−08−31. 基金项目：国家自然科学基金项目 (61973116). 通信作者：尹诗. E-mail：yinshi502@163.com. 第 16 卷第 6 期 智 能 系 统 学 报 Vol.16 No.6 2021 年 11 月 CAAI Transactions on Intelligent Systems Nov. 2021