CEEMDAN联合自适应小波阈值算法的GA-BP风电发电机故障预测  

CEEMDAN joint adaptive wavelet thresholding algorithm for GA-BP wind turbine fault prediction

作　　者：肖成 曹万鹏 褚越强 杨政琨 王佳兴 Xiao Cheng;Cao Wanpeng;Chu Yueqiang;Yang Zhengkun;Wang Jiaxing(School of Electronics and Control Engineering,North China Institute of Aerospace Engineering,Langfang 065000,China;China Suntien Green Energy Corporation Limited,Huaian 223001,China)

机构地区：[1]北华航天工业学院电子与控制工程学院,河北廊坊065000 [2]新天绿色能源股份有限公司,江苏淮安223001

出　　处：《可再生能源》2024年第10期1332-1340,共9页Renewable Energy Resources

基　　金：河北省教育厅重点项目(ZD2022089);北华航天工业学院博士基金项目(BKY-2023-03);北华航天工业学院校重点项目(ZD-2022-03)。

摘　　要：发电机是风电系统中重要的核心部件,为了提高风电机组的稳定、高效运行,对风电机组发电机的故障预测十分必要。文章围绕风电系统发电机机侧轴承温度超限故障预测的问题,考虑到所采集的故障特征信号具有较大噪声的特点,引入自适应完备噪声经验模态分解(CEEMDAN)联合自适应小波阈值去噪的方法实现信号有效去噪,同时结合GA-BP神经网络建立故障预测模型。通过与BP神经网络、GA-BP神经网络对比预测指标、误差指标和预测效果图形,验证了所提算法可以获得较好的预测效果。误差指标和预测效果均有提升,对提前15 d风电系统发电机故障预测的准确率达到了92.98%。Generator is an important core component in wind power system,in order to improve the stable and efficient operation of wind turbine,the fault prediction of wind turbine generator is necessary.Focusing on the problem of generator machine-side bearing temperature overrun fault prediction in wind power system,this paper takes into account that the collected fault characteristic signal is characterized by large noise,introduces CEEMDAN joint adaptive wavelet threshold denoising method to realize effective denoising of the signal,and at the same time establishes a fault prediction model by combining GA-BP neural network.By comparing the prediction indexes,error indexes and prediction effect graphs with BP neural network and GA-BP neural network,it is verified that the proposed algorithm can obtain better prediction effect.The error index and prediction effect are improved,and the accuracy of the prediction of generator failure of wind power system 15 days in advance reaches 92.98%.

关 键 词：风电系统 发电机故障 故障预测 CEEMDAN GA-BP神经网络 

分 类 号：TK81[动力工程及工程热物理—流体机械及工程]