风力发电机组的非侵入式机械故障检测与诊  

Non-invasive Mechanical Fault Detection and Diagnosis of Wind Turbine

作　　者：李玥 杨竣辉[2] LI Yue;YANG Junhui(Ganzhou Teachers College,Ganzhou Jiangxi 341000,China;Jiangxi University of Technology,School of Information Engineering,Ganzhou Jiangxi 341000,China)

机构地区：[1]赣州师范高等专科学校,江西赣州341000 [2]江西理工大学信息工程学院,江西赣州341000

出　　处：《机械设计与研究》2023年第6期248-254,共7页Machine Design And Research

基　　金：国家自然科学基金资助项目(61273328,61805053);江西省教育厅科学技术研究项目(GJJ181188);江西省高等学校教学改革研究课题(JXJG-21-46-1)。

摘　　要：为在不使用机械传感器的前提下,有效对风力发电机组的机械故障进行检测与诊断,提出一种风力发电机组的非侵入式机械故障检测与诊断策略,借助改进模糊Q学习分类器实现5种机械故障类型的精确诊断。首先对风力发电机组的模型搭建和不平衡故障的构建方式进行介绍。接着提出用于风力发电机组的非侵入式机械故障检测与诊断策略,通过经验模态分解(EMD)技术对机械故障工况进行特征提取后,使用决策树J48算法筛选出最具影响力的输入(MII),并介绍改进模糊Q学习分类器的设计方法。最后利用实验平台得到的数据集用于诊断策略的训练和效果验证,实验结果表明所提出的机械故障诊断策略可以实现99.47%的平均正确率,明显高于作为对照的支持向量机、神经网络和传统模糊Q学习策略。In order to effectively detect and diagnose the mechanical faults of wind turbine generators without using mechanical sensors,a non-invasive mechanical fault detection and diagnosis strategy for wind turbine generators is proposed,and accurate diagnosis of five types of mechanical faults is achieved by using an improved fuzzy Q learning classifier.Firstly,the construction of wind turbine model and the construction of unbalanced fault are briefly analyzed.Then,a non-invasive mechanical fault detection and diagnosis strategy for wind turbine generators is proposed.After feature extraction of mechanical fault conditions through EMD technology,the decision tree J48 algorithm is used to screen the most influential input(MII),and the design method of the proposed improved fuzzy Q learning classifier is derived.Finally,the existing data sets in the laboratory are used as training and effect verification.The experimental results show that the proposed mechanical fault diagnosis strategy can achieve an average accuracy of 99.47%,which is significantly higher than the control support vector machine,neural network and traditional fuzzy Q learning strategies.

关 键 词：风力发电机组 非侵入式 机械故障 改进模糊Q学习 故障诊断 

分 类 号：TH132.41[机械工程—机械制造及自动化]