基于状态切分与时间卷积网络的风电齿轮箱系统温度异常故障预警  

Fault Early Warning of Wind Turbine Gearbox System Temperature Anomaly Based on State Segmentation and Temporal Convolutional Network

作　　者：郭涵涛 滕伟[1] 高晨 童博[2] 彭迪康 柳亦兵[1] Han-tao Guo;Wei Teng;Chen Gao;Bo Tong;Di-kang Peng;Yi-bing Liu(Key Laboratory of Power Station Energy Transfer Conversion and System of Ministry of Education,North China Electric Power University;Xi’an Thermal Power Research Institute Co.,Ltd.)

机构地区：[1]华北电力大学电站能量传递转化与系统教育部重点实验室 [2]西安热工研究院有限公司

出　　处：《风机技术》2023年第4期65-71,共7页Chinese Journal of Turbomachinery

基　　金：自然科学基金项目(51775186);中国华能集团总部科技项目(HNKJ20-H72)。

摘　　要：由于风电齿轮箱结构复杂、运行工况多变,采用监控与数据采集系统(SCADA)数据对齿轮箱早期故障预警的精度不足。本文提出考虑风电齿轮箱润滑冷却状态及运行工况的状态切分方法,基于齿轮箱润滑冷却系统运行原理,从原始运行数据中选取相关参数,采用统计分析和聚类方法将齿轮箱系统时间序列数据进行运行策略分类切分,构建齿轮箱运行状态判断模型;提出采用时间卷积神经网络训练不同运行策略下的齿轮箱温度预测模型,并实时判断运行状态,选取对应运行策略温度预测模型,估计齿轮箱温度,通过与实际值之间的残差实现齿轮箱故障预警。实际案例表明,本文所提出的方法可以提高模型精度,能够有效预警风电齿轮箱系统的早期故障。The structure of wind turbine gearboxes is complex,and the operating conditions of wind turbines are changeable.The accuracy of early fault warning using SCADA data is insufficient.Aiming at the problem of frequent failure and the difficulty of early fault monitoring of wind turbine,a method base on state segmentation and temporal convolutional network is proposed.Firstly,based on the operating principle of the gearbox lubrication and cooling system,relevant parameters are selected from the original operating data.The time series data of the gearbox system are classified and segmented by statistical analysis and clustering methods to build the judgment model of the gearbox operating strategy.Secondly,temporal convolutional network is used to train the gearbox temperature prediction model under different operation strategies.Finally,after judging the current running strategy,temperature prediction models under the current operation strategy are selected to estimate the gearbox temperature and monitor the gearbox fault.The actual case shows that the proposed method can improve the accuracy of the temperature prediction model.The gearbox early fault wind turbine can be effectively detected by using the method.

关 键 词：风电机组 齿轮箱系统 故障预警 状态切分 

分 类 号：TK83[动力工程及工程热物理—流体机械及工程]