摘要
针对风电机组存在的典型故障进行了归纳,选取某风场风电机组故障频次较高的变流系统、发电机系统、变桨系统、辅助电源系统故障数据和非故障数据进行故障诊断研究,分别采用极限学习机(ELM)、最小二乘支持向量机(SVM)、核极限学习机(KELM)和鲸鱼群优化算法(WOA)的WOA-KELM算法建立了故障诊断模型,同时采用拉普拉斯分数对模型特征变量重要程度进行排序和选取,WOA-KELM算法通过优化KELM算法的正则化参数C与核参数γ取得了更好的诊断效果。研究表明:不同样本数量下4种算法4对非故障类型的诊断准确率均为100%;采用拉普拉斯分数对WOA-KELM算法进行特征筛选后测试样本的平均诊断准确率从88.0%提高到93.2%;WOA-KELM算法在样本数量为250~500内进行特征筛选后的诊断准确率达到最大值96.0%。这证明该模型可以有效实现风电机组的故障诊断,为现场运维人员提供指导与参考。
The typical faults of wind turbines are summarized.The fault data and non-fault data of converter system,generator system,variable propeller system and auxiliary power system with high fault frequency of wind turbines in a wind farm are selected for fault diagnosis research.The fault diagnosis model is established by ELM,SVM,KELM and WOA-KELM algorithms respectively.At the same time,Laplacian scores are used to sort and select the importance degree of model characteristic variables.WOA-KELM algorithm achieves better diagnostic effect by optimizing the regularization parameter C and kernel parameterγ of KELM algorithm.The results show that,the diagnostic accuracy of the four algorithms for non-fault types is 100%under different sample numbers.The average diagnostic accuracy of WOA-KELM algorithm improves from 88.0%to 93.2%after feature screening by using Laplace scores.In the range of 250~500 samples,the diagnostic accuracy of WOA-KELM algorithm reaches the maximum of 96.0%after feature screening.It is proved that this model can effectively realize the fault diagnosis of wind turbine,and provide guidance and reference for field operation and maintenance personnel.
作者
安留明
沙德生
张庆
李芊
刘潇波
张鑫赟
AN Liuming;SHA Desheng;ZHANG Qing;LI Qian;LIU Xiaobo;ZHANG Xinyun(China Huaneng Clean Energy Research Institute Co.,Ltd.,Beijing 102209,China)
出处
《热力发电》
CAS
CSCD
北大核心
2023年第12期131-139,共9页
Thermal Power Generation
基金
中国华能集团清洁能源技术研究院有限公司研究与开发基金项目(QNYJJ22-18)。
