摘要
针对复杂多变的工作环境给风电机组状态监测带来的挑战,提出了一种基于深度学习和注意力机制组合的状态监测与健康评估方法。首先,将风电机组数据采集与监控(supervisory control and data acquisition,简称SCADA)系统数据进行预处理;其次,将卷积神经网络(convolutional neural networks,简称CNN)和长短期记忆网络(long short-term memory,简称LSTM)相结合提取数据的时空特征,并引入注意力机制(Attention)为LSTM分配相应的权重;然后,利用指数加权移动平均来设置阈值,通过分析均方根误差实现风电机组的状态监测;最后,通过实例对风电机组的主轴承、发电机定子和叶片变桨电机状态进行监测分析和健康评估,验证该方法的有效性。
The complex and dynamic working environment brings challenges to the condition monitoring of wind turbines.This paper presents a condition monitoring and health assessment method based on convolutional neural networks and long short-term memory(CNN-LSTM) with attention.The method first preprocesses the data of the supervisory control and data acquisition(SCADA) system of wind turbines and then combines the convolutional neural networks(CNN) and long short-term memory(LSTM) networks to fully extract the timespace characteristics of the data itself.At the same time,an attention mechanism is introduced to assign corresponding weight to LSTM.The exponentially weighted moving average is used to set the threshold value,and the root mean square error is analyzed to realize the status monitoring of wind turbines.Finally,the monitoring analysis and health assessment of the main bearing,generator stator and blade pitch motor of the wind turbine is carried out through an example to verify the effectiveness of the method.
作者
朱岸锋
赵前程
周凌
杨天龙
阳雪兵
ZHU Anfeng;ZHAO Qiancheng;ZHOU Ling;YANG Tianlong;YANG Xuebing(School of Mechanical Engineering,Hunan University of Science and Technology Xiangtan,411201,China;School of Electrical and Information Engineering,Hunan University of Technology Zhuzhou,412002,China;XEMC Windpower Co.,Ltd.Xiangtan,411102,China)
出处
《振动.测试与诊断》
北大核心
2025年第2期256-263,409,共9页
Journal of Vibration,Measurement & Diagnosis
基金
国家重点研发计划资助项目(2022YFF0608700)
国家自然科学基金资助项目(51875199)
湖南省教育厅青年资助项目(22B0590)。
关键词
风电机组
数据采集与监控系统
神经网络
状态监测
健康评估
wind turbine
supervisory control and data acquisition system
neural network
condition monitoring
health assessment
