作者：王东林¹, 吕丽霞¹, 王梓齐¹, 陈颖², 李晓宁³

作者单位：1. 华北电力大学控制与计算机工程学院，河北 保定071000;
2. 火箭军指挥学院，湖北 武汉430012;
3. 大唐武安发电有限公司，河北 邯郸 056300

关键词：风电机组;状态监测;深度自编码网络;高斯混合模型;工况辨识

摘要：

针对大型风电机组运行工况多变、数据量大的特点，提出一种将高斯混合模型（GMM）与深度自编码网络（DAE）相结合的风电机组齿轮箱状态监测方法。首先，基于GMM对风电机组运行工况进行辨识；然后，在各个子工况空间下，基于DAE建立正常运行状态下的齿轮箱油池温度模型，得到多工况阈值；最后，对DAE模型的重构误差进行分析，结合多工况阈值构建健康指数，实现风电机组齿轮箱的状态监测。以某台2 MW风电机组为实例进行验证，结果表明，该方法能够提前7天预警齿轮箱油池温度过高的故障；相对于基于DAE状态监测方法，在不影响在线监测时效性的情况下，该文所提方法能够提前约8 h预警潜在故障。

Condition monitoring of wind turbine gearbox based on GMM operational condition identification and DAE
WANG Donglin¹, Lü Lixia¹, WANG Ziqi¹, CHEN Ying², LI Xiaoning³
1. School of Control and Computer Engineering, North China Electric Power University, Baoding 071000, China;
2. The Rocket Force Command College, Wuhan 430012, China;
3. Datang Wuan Power Generation Co., Ltd., Handan 056300, China
Abstract: In view of the variable operational conditions of large wind turbines and large amount of data, a wind turbine gearbox condition monitoring method based on Gaussian mixture model (GMM) and deep auto-encoder network (DAE) is proposed. Firstly, GMM is used to identify the operational conditions of wind turbine. Then, under each sub-condition space, DAE is used to establish the temperature models of the gearbox oil to obtain the multi-condition warning threshold. Finally, the reconstruction error of gearbox oil based on the DAE model is analyzed. The health index is constructed based on reconstruction error and multi-condition threshold to realize the condition monitoring. A 2 MW wind turbine is used as an example for verification. The results show that the proposed method can warn of potential faults 7 days in advance. Being compared with the method only based on DAE, without affecting the timeliness of online monitoring, this method can warn 8 h in advance.
Keywords: wind turbine;condition monitoring;deep auto-encoder network;Gaussian mixture model;operational condition identification
2021, 47(4):89-95  收稿日期: 2020-06-24;收到修改稿日期: 2020-07-22
基金项目: 北京市自然科学基金资助项目(4182061)；中央高校基本科研业务费专项资金资助(2020JG006，2020MS117)
作者简介: 王东林（1994-），男，河北唐山市人，硕士研究生，专业方向为风电机组状态监测
参考文献
[1] LIN Y, TU L, LIU H, et al. Fault analysis of wind turbines in China[J]. Renewable & Sustainable Energy Reviews, 2016, 55: 482-490
[2] 乔福宇, 马良玉, 马永光. 基于功率曲线分析与神经网络的风电机组故障预警方法[J]. 中国测试, 2020, 46(8): 44-50
[3] HELBING G, RITTER M. Deep learning for fault detection in wind turbines[J]. Renewable & Sustainable Energy Reviews, 2018, 98: 189-198
[4] 王洪斌, 王红, 何群, 等. 基于深度信念网络的风机主轴承状态监测方法[J]. 中国机械工程, 2018, 29(8): 948-953
[5] KONG Z, TANG B, Deng L, et al. Condition monitoringof wind turbines based on spatio-temporal fusion of SCADA data by convolutional neural networks and gated recurrent units[J]. Renewable Energy, 2020, 146: 760-768
[6] ZHAOH, LIU H, HU W, et al. Anomaly detection and fault analysis of wind turbine components based on deep learning network[J]. Renewable Energy, 2018, 127(11): 825-834
[7] 董玉亮, 李亚琼, 曹海斌, 等. 基于运行工况辨识的风电机组健康状态实时评价方法[J]. 中国电机工程学报, 2013, 33(11): 88-95
[8] 刘长良, 闫萧. 基于工况辨识的风电机组齿轮箱状态监测[J]. 热能动力工程, 2016, 31(7): 41-46, 133
[9] 王梓齐, 刘长良, 刘帅. 基于集成NSET和模糊软聚类的风电机组齿轮箱状态监测[J]. 仪器仪表学报, 2019, 40(7): 138-146
[10] WANG H, WANG H B, JIANG G Q, et al. Early fault detection of wind turbines based on operational condition clustering and optimized deep belief network modeling[J]. Energies, 2019, 12(6): 984-1006
[11] 尹诗, 侯国莲, 于晓东, 等. 基于Bi-RNN的风电机组主轴承温度预警方法研究[J]. 郑州大学学报(工学版), 2019, 40(5): 45-51
[12] 刘帅, 刘长良, 甄成刚, 等. 基于群体多维相似性的风机齿轮箱预警策略[J]. 仪器仪表学报, 2018, 39(1): 180-189
[13] HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computa, 2014, 18(7): 1527-1554
[14] HINTON G E, SALAKHUTDINOV R. Reducing the dimensionality of data with neural networks.[J]. Science, 2006, 313(5786): 504-507
[15] 尤波, 李忠杰, 黄玲. 基于改进型BP神经网络的手部动作识别[J]. 智能系统学报, 2018, 13(5): 848-854