基于改进GRU的海上风电机组齿轮箱故障诊断  

Fault Diagnosis of Offshore Wind Turbine Gearbox Based on Improved GRU

作　　者：徐进 汤海宁 丁显 XU Jin;TANG Haining;DING Xian(China Green Development Investment Group Co.,Ltd.,Beijing 100020,China;Duchengweiye Group Co.,Ltd.,Beijing 100032,China;NARI-TECH Nanjing Control Systems Ltd.,Nanjing 211106,China)

机构地区：[1]中国绿发投资集团有限公司,北京100020 [2]都城伟业集团有限公司,北京100032 [3]国电南瑞南京控制系统有限公司,南京211106

出　　处：《船舶工程》2022年第9期167-173,182,共8页Ship Engineering

摘　　要：为提高风电机组中齿轮箱故障诊断的准确性,提出一种在门控循环单元(GRU)神经网络中引入注意力机制进行风电机组齿轮箱油温预测的方法,充分挖掘数据参量中的关联关系,以提高模型预测的准确性。针对模型预测的残差,采用一种基于粒子群算法的自适应阈值方案进一步实现对风电机组齿轮箱油温异常的诊断,引入自适应阈值能提高模型诊断的灵敏度。针对江苏某海上风电场实际运行的SCADA数据进行仿真分析和验证,结果证明采用该方法能较早地判断出风电机组齿轮箱的异常状态。To improve the accuracy of gearbox fault diagnosis.A method is presented to predict the oil temperature of wind turbine gearbox by introducing attention mechanism into GRU neural network,which fully excavates the correlation relationship in data parameters to improve the accuracy of model prediction.Furthermore,an adaptive threshold scheme based on particle swarm optimization algorithm is used to further diagnose the abnormal oil temperature of wind turbine gearbox.The introduction of adaptive threshold improves the diagnostic sensitivity of the model.Through the simulation analysis and verification of SCADA data of an offshore wind farm in Jiangsu Province,it is proved that this method can judge the abnormal state of wind turbine gearbox earlier.

关 键 词：海上风电 齿轮箱油温预测 故障诊断 注意力机制 门控循环单元 

分 类 号：U665.26[交通运输工程—船舶及航道工程]