文章摘要
毕扬,郑波,张亚武,朱溪,江亚兰,李超顺.基于MIC与BiGRU的水电机组振动趋势预测[J].水利学报,2021,52(5):612-621,632
基于MIC与BiGRU的水电机组振动趋势预测
Vibration trend prediction of hydroelectric generating unit based on MIC and BiGRU
投稿时间:2020-07-22  
DOI:10.13243/j.cnki.slxb.20200553
中文关键词: 最大信息系数法  BiGRU  小波阈值去噪  信号处理  特征选择  趋势预测
英文关键词: maximal information coefficient (MIC)  bidirectional gate recurrent unit (BiGRU)  wavelet threshold de-noising (WTD)  signal processing  feature selection  trend prediction
基金项目:国家自然科学基金项目(51879111);武汉市应用基础前沿专项(2018010401011269);湖北省自然科学基金项目(2019CFA068)
作者单位E-mail
毕扬 浙江仙居抽水蓄能有限公司, 浙江 仙居 317300  
郑波 中国电建集团华东勘测设计研究院有限公司, 浙江 杭州 311122  
张亚武 国网新源控股有限公司, 北京 100761  
朱溪 浙江仙居抽水蓄能有限公司, 浙江 仙居 317300  
江亚兰 华中科技大学 水电与数字化工程学院, 湖北 武汉 430074  
李超顺 华中科技大学 水电与数字化工程学院, 湖北 武汉 430074 csli@hust.edu.cn 
摘要点击次数: 1558
全文下载次数: 763
中文摘要:
      为提高水电机组振动趋势预测的准确率,本研究提出了一种基于最大信息系数(MIC)与双边门控循环神经网络(BiGRU)的水电机组振动趋势预测模型。首先,预处理阶段采用小波系数阈值去噪(WTD)方法对历史振动信号数据进行降噪处理以消除强背景噪声的干扰,并将振动信号划分为多个训练样本以改善算法的鲁棒性;其次考虑水力、电气与机械不平衡力因素的影响,基于MIC对与振动信号关联性强的状态参数进行特征选择作为模型的参考输入;再采用BiGRU网络建立振动信号预测模型,进行超前多步的振动信号趋势预测;最后利用训练好的模型对在线获取的振动数据进行实时预测。为评估模型的预测性能,本研究采集某抽水蓄能水电站的振动监测数据进行多组对比实验,验证了所提方法具有较好的预测能力和泛化能力,适用于水电机组振动的趋势预测。
英文摘要:
      To improve the accuracy of vibration trend-prediction for hydropower unit,a novel trend-prediction approach is proposed based on maximal information coefficient (MIC) and bidirectional gate recurrent unit (BiGRU) network. Initially, wavelet threshold de-noising (WTD) method is applied to reduce the strong noise of historical vibration data. Then, the MIC algorithm is used to select the strongly correlated status parameters,which influenced by hydraulic,electrical factors and mechanical unbalanced force. Then, BiGRU network is employed to establish the vibration trend-prediction model, and the trend prediction of vibration signals is carried out with multistep in advance. The trend-prediction of vibration signal obtained online is predicted by the trained model. In order to evaluate the prediction performance of the model,the vibration monitoring data of a pumped storage hydropower station are collected for multiple groups of comparative experiments, which proves that the proposed method has good prediction ability and generalization ability,which is suitable for the trend prediction of vibration of hydropower units.
查看全文   查看/发表评论  下载PDF阅读器
关闭