作者：张航¹, 史兆培¹, 束垠¹, 张子瑞¹, 宋志强¹, 许昌²

作者单位：1. 中国三峡新能源（集团）股份有限公司，北京 101100;
2. 河海大学能源与电气学院，江苏 南京 211100

关键词：风电机组;故障诊断;机器学习;特征提取;贝叶斯优化

摘要：

随着风电机组大规模的并网运行，风电机组的故障诊断逐渐成为行业的研究热点。风电机组故障的及时发现与处理，能够保障机组的安全稳定运行，提升风电场经济效益。通过SCADA运行数据进行风电机组故障诊断是一种重要的故障诊断诊断方式，文章从故障诊断的特征提取及故障诊断模型构建角度出发，提出应用随机森林袋外估计（OOB）进行特征选择的特征提取方法和改进参数优化机器学习算法(BO-LightGBM)的风电机组故障诊断模型，提高基于数据驱动的风电机组故障预测的精度。通过风电场实际运行数据对所提故障诊断方法进行验证，结果证明该模型对于不同类型的故障均有92%以上的预测准确性，表明该模型对风电机组故障诊断具有较好的适用性。

Fault diagnosis method of wind turbine based on OOB-BO-LightGBM
ZHANG Hang¹, SHI Zhaopei¹, SHU Yin¹, ZHANG Zirui¹, SONG Zhiqiang¹, XU Chang²
1. China Three Gorges Renewables (Group) Co., Ltd., Beijing 101100, China;
2. College of Energy and Electrical Engineering, Hohai University, Nanjing 211100, China
Abstract: With the large-scale grid-connected operation of wind turbines, fault diagnosis of wind turbines has gradually become a research hotspot in the industry. The timely detection and treatment of wind turbine faults can ensure the safe and stable operation of the wind turbines and improve the economic benefits of the wind farm. It’s an important way for fault diagnosis of wind turbines by the analysis of SCADA data. From the perspective of fault diagnosis feature extraction and fault diagnosis model construction, this paper proposes a feature extraction method using random forest out-of-bag estimation for feature selection and a wind turbine fault diagnosis model with improved parameter optimization machine learning algorithms to improve data-driven wind power The accuracy of unit failure prediction. The proposed fault diagnosis method is verified by actual wind farm operating data. The results prove that the model has a predictive accuracy of more than 92% for different types of faults, indicating that the model has good applicability for fault diagnosis of wind turbines.
Keywords: wind turbine;fault diagnosis;machine learning;feature extraction;Bayesian optimization
2024, 50(4):83-89  收稿日期: 2021-12-13;收到修改稿日期: 2022-04-15
基金项目: 国家自然基金委雅砻江联合基金资助项目（U1865101）；中国三峡新能源（集团）股份有限公司科研项目资助（34009024）；政府间国际科技创新合作重点专项（2019YFE0104800）
作者简介: 张航(1991-),男,江苏镇江市人,硕士,主要研究方向为风电场故障诊断和控制。
参考文献
[1] CHU S, MAJUMDAR A. Opportunities and challenges for a sustainable energy future[J]. Nature, 2012, 488(7411): 294-303
[2] 沈学华, 周志华, 吴建鑫, 等. Boosting 和 Bagging 综述[J]. 计算机工程与应用, 2000(12): 31-32
SHEN X H, ZHOU Z H, WU J X, et al. Boosting and bagging review[J]. Computer Engineering and Applications, 2000(12): 31-32
[3] BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32
[4] 郭莹莹, 张磊, 肖成, 等. 基于改进深度森林算法的风电机组故障诊断技术研究[J]. 可再生能源, 2019, 37(11): 1720-1725
GUO Y Y, ZHANG L, XIAO C, et al. Research on wind turbine fault diagnosis technology based on improved deep forest algorithm[J]. Renewable Energy, 2019, 37(11): 1720-1725
[5] 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
LI Y, AN P, LI Z Q, et al. Research on fault warning level based on NSET and information entropy[J]. China Measurement & Test, 2020, 46(7): 153-158
[6] 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
QIAO F Y, MA L Y, MA Y G. Wind turbine fault early warning method based on power curve analysis and neural network[J]. China Measurement & Test, 2020, 46(8): 44-50
[7] 韩中合, 赵文波, 朱霄珣, 等．基于KL-CEEMD的风机传动系统故障诊断方法研究[J]．中国测试,2022,48(5):88-95
HAN Z H, ZHAO W B, ZHU X J, et al. Research on fault diagnosis method of fan drive system based on KL-CEEMD [J]. China Measurement ＆ Test, 2022, 48 (5): 88-95.
[8] 钱力扬. 基于随机森林和XGBoost的大型风力机故障诊断方法研究[D]. 杭州: 浙江大学, 2018.
QIAN L Y. Research on fault diagnosis method for large wind turbine based on random forest and XGBoost [D]. Hangzhou: Zhejiang University, 2018.
[9] FRIEDMAN J H. Greedy function approximation: A gradient boosting machine[J]. Annals of Statics, 2001, 29(5): 1189-1232
[10] CARL E R. Gaussian processes in machine learning[C]. Machine Learning Summer School Conference, 2003.
[11] HOFFMAN M W, BROCHU E, DE F N. Portfolio allocation for bayesian optimization[C]. Proceedings of the 27th Conference on Uncertainty in Artificial Intelligence. AUAI Press, 2011: 327-336.
[12] 李俊卿, 李秋佳, 石天宇, 等. 基于数据挖掘的风电功率预测特征选择方法[J]. 电测与仪表, 2019, 56(10): 87-92
LI J Q, LI Q J, SHI T Y, et al. A feature selection method for wind power prediction based on data mining[J]. Electrical Measurement and Instrumentation, 2019, 56(10): 87-92