文章摘要
基于振动信号的水电机组状态劣化在线评估方法研究
Research on On-line Evaluation Method of Hydropower Unit Condition Degradation Based on Vibration Signal
投稿时间:2020-05-01  修订日期:2020-09-03
DOI:
中文关键词: 水电机组  劣化评估  振动信号  检测指数  最小二乘支持向量机  小波奇异值
英文关键词: hydropower unit  degradation assessment  vibration signal  detection index  least squares support vector machine  wavelet singular value
基金项目:国家自然科学基金项目(51979204); 国家自然科学基金项目(51379160);
作者单位E-mail
刘东 武汉大学 水资源与水电工程科学国家重点实验室 LewistWHU@163.com 
赖旭 武汉大学 水资源与水电工程科学国家重点实验室 laixu@whu.edu.cn 
胡晓 武汉大学 水力机械过渡过程教育部重点实验室  
肖志怀 武汉大学 水力机械过渡过程教育部重点实验室  
摘要点击次数: 1063
全文下载次数: 0
中文摘要:
      实现水电机组状态劣化评估和故障预警是行业研究的热点。论文提出了一种结合时域与频域特征的机组劣化在线评估方法。一方面,利用检测指数确定振动信号中对机组运行状态最为敏感的时域特征,以机组健康状态下工况参数(水头、开度等)和检测指数筛选的振动信号时域特征为健康样本,利用最小二乘支持向量机构建机组状态健康模型。利用该模型,以实时工况参数为输入,在线预测对应工况下机组振动信号时域特征健康值,计算健康值与实际值之间的相对误差,作为评估机组劣化程度的时域劣化指标。另一方面,利用小波变换与奇异值理论对振动信号进行分解,提取健康状态下机组振动信号奇异值特征向量并得到健康聚类中心,实时计算实测信号奇异值特征向量与健康聚类中心之间的相对欧式距离,作为频域劣化指标。结合时域和频域劣化指标,在线计算综合劣化指标评估当前时刻机组劣化程度。结合实际机组运行案例,验证了该模型的有效性和实用性。
英文摘要:
      Realizing the deterioration assessment and the faults warning of hydropower units is hot topics in the industry. The paper proposes an online evaluation method for unit degradation based on the combination of time and frequency domain features. On the one hand, the detection index is used to determine the time domain characteristics of the vibration signal that are most sensitive to the operating state of the unit. The operating parameter X (head, opening, etc.) of the unit in health state and the time domain characteristics Y of the vibration signal selected by the detection index are selected as healthy sample, and the least squares support vector machine is used to construct the unit state health model Y=f(X). The real-time operating condition parameters are inputted into this model, and the health value of the unit's vibration signal time-domain characteristic under corresponding operating conditions is predicted online. The relative error between the health value and the actual value is calculated as a time-domain degradation index. On the other hand, the wavelet transform and singular value theory are used to decompose the vibration signal to extract the singular value feature vector of the unit's vibration signal under healthy state and obtain the health clustering center. The relative Euclidean distance is used as an index of frequency domain degradation. Combined with the time and frequency domain degradation indicators, an online calculation of comprehensive degradation indicators is performed to assess the degree of unit degradation at the current moment. Combined with actual unit operation cases, the effectiveness and practicability of the model are verified.
  查看/发表评论  下载PDF阅读器
关闭