作者：于航, 尹诗

作者单位：中能电力科技开发有限公司，北京 100034

关键词：门控循环单元神经网络;梯度提升迭代决策树;状态监测;LightGBM;风电机组

摘要：

针对风电机组发电机前轴承运行过程中早期异常状态识别的问题，提出一种基于GRU-LightGBM的风电机组发电机前轴承状态监测方法。首先，通过梯度提升迭代决策树(gradient boosting decision tree, GBDT)算法分析SCADA历史数据，提取与风电机组发电机前轴承温度特征相关性强的特征变量。然后，采用门控递归单元(gated recurrent unit, GRU)神经网络建立风电机组发电机前轴承温度预测模型并计算残差特征。最后，采用LightGBM算法建立故障决策模型进行状态监测。实验研究表明：该方法能有效识别发电机前轴承运行状态，能够在故障发生前一个月识别风电机组发电机前轴承的异常运行，对风电机组设备运行维修和早期故障预警具有借鉴意义。

Wind turbine generator front bearing condition monitoring based on GRU-LightGBM algorithm
YU Hang, YIN Shi
Zhongneng Power Technology Development Co., Ltd., Beijing 100034, China
Abstract: Aiming at the problem of early identification of abnormal conditions during the operation of front bearing of the wind turbine generator, a GRU-LightGBM wind turbine generator front bearing state monitoring method is proposed. First, the historical data of SCADA is analyzed through the gradient boosting decision tree algorithm, and the characteristic variables that have strong correlation with the temperature characteristics of front bearing of the wind turbine generator are extracted. Then, a gated recurrent unit (GRU) neural network is used to establish wind turbine generator front bearing temperature prediction model, and calculate the characteristics of the residual value. Finally, using the LightGBM algorithm to establish fault decision model for state detection. Experimental research shows that this method can effectively identify the running state of front bearing of the generator, and can identify the abnormal running state of front bearing of wind turbine generator one month before fault occurs. It has reference significance for wind turbine equipment operation maintenance and early fault early warning.
Keywords: gated recurrent unit networks;gradient boosting decision tree;condition monitoring;LightGBM;wind turbine
2022, 48(9):105-111  收稿日期: 2021-07-19;收到修改稿日期: 2021-09-21
基金项目: 国家自然科学基金(61973116)
作者简介: 于航（1976-），男，吉林白城市人，高级工程师，硕士，研究方向为新能源故障预警、新能源大数据分析等
参考文献
[1] 刘帅, 刘长良, 甄成刚. 基于数据分类重建的风电机组故障预警方法[J]. 仪器仪表学报, 2019, 40(8): 1-11
[2] 陈雪峰, 郭艳婕, 许才彬, 等. 风电装备故障诊断与健康监测研究综述[J]. 中国机械工程, 2020, 31(2): 175-189
[3] 向健平, 凌永志, 詹俊, 等. 基于SCADA系统的风电机组主轴承故障预警方法[J]. 电力科学与技术学报, 2019, 34(3): 223-228
[4] 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
[5] 刘帅, 刘长良, 曾华清. 基于核极限学习机的风电机组齿轮箱故障预警研究[J]. 中国测试, 2019, 45(2): 121-127
[6] 李洋, 安平, 李志强, 等. 基于NSET与信息熵的故障预警等级研究[J]. 中国测试, 2020, 46(7): 153-158
[7] 孙鹤旭, 孙泽贤, 张靖轩. 基于并行模糊C-均值聚类的风电机组发电机故障诊断研究[J]. 太阳能学报, 2020, 41(3): 8-14
[8] 马宏忠, 李思源. 双馈风力发电机转子绕组不平衡故障诊断[J]. 电机与控制学报, 2020, 24(2): 138-143
[9] 杨文珺, 何婷, 王贵军. 风力发电机组的故障及维护策略[J]. 机械研究与应用, 2019, 32(5): 190-193,198
[10] CHO K, MERRIENBOER B V, GULCEHRE C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[J]. Computer Science, 2014, 6(3): 1-15
[11] CHEN T, GUESTRIN C. XGBoost: a scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016.
[12] KE G, MENG Q, FINLEY T, et al. LightGBM: a highly efficient gradient boosting decision tree[C]//Advances in Neural Information Processing Systems, 2017.
[13] CHUNG J, GULCEHRE C, CHO K H, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[J]. Eprint Arxiv, 2014(12): 1-9
[14] 周挺, 杨军, 周强明, 等. 基于改进LightGBM的电力系统暂态稳定评估方法[J]. 电网技术, 2019, 43(6): 1931-1940