基于瞬时轨迹特征图像和条件对抗生成网络的水电机组轴系劣化评估  

Degradation Assessment of Shaft System of Hydropower Unit Based on Instantaneous Orbit Feature Image and Conditional Generative Adversarial Network

作　　者：王义国 常辉[2] 陈东君 吴一凡 李超顺[4] WANG Yi-guo;CHANG Hui;CHEN Dong-jun;WU Yi-fan;LI Chao-shun(Guangdong Energy Group Co.,Ltd.,Guangzhou 510630,China;Tianshengqiao First-class Hydropower Development Co.,Ltd.,Xingyi 562400,China;Guangdong Yudean Changhu Power Generation Co.,Ltd.,Yingde 513021,China;School of Civil and Hydraulic Engineering,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]广东能源集团有限公司,广东广州510630 [2]天生桥一级水电开发有限责任公司,广东兴义562400 [3]广东粤电长湖发电有限责任公司,广东英德513021 [4]华中科技大学土木与水利工程学院,湖北武汉430074

出　　处：《水电能源科学》2023年第6期166-170,共5页Water Resources and Power

基　　金：国家自然科学基金面上项目(51879111)。

摘　　要：水电机组轴系对机组稳定性影响显著,针对机组轴系的劣化评估能够直观反映机组运行状态。因此,提出了一种利用瞬时轨迹特征图像和条件对抗生成网络的机组轴系劣化评估方法。先将各支承面反映振动轨迹的垂直信号构造为复信号,使用多变量复信号变分模态分解方法将信号分解并提取椭圆瞬时轨迹离心率、正反向进动分量等瞬时轨迹特征,构造出多个轴承面的特征图像。使用条件对抗生成网络构建健康模型,将机组工况参数和白噪声作为输入,分别拟合出每个轴承面在不同工况中健康状态下特征图像的分布,利用真实图像与健康图像的差异构建健康指标。利用遗传算法对多个轴承面的评价指标权重进行寻优,降低健康区间指标波动性,获取综合轴系劣化曲线。在水电机组轴系数据中进行试验,证明了所提方法的有效性。The shaft system of hydropower unit has a significant impact on the stability of the unit.The degradation assessment of the shaft system can visually reflect the operating condition of the unit.This paper presents a method for assessing the degradation of the unit’s shaft system using instantaneous orbit feature image and conditional adversarial generative network(CGAN).Firstly,the vertical signals of each bearing were constructed as a complex signal,and the multivariate complex variational mode decomposition(MCVMD)method was used to process the signal and extract the instantaneous orbit features to construct the instantaneous orbit feature images.CGAN was used to construct the health model.The health model can fit the distribution of feature images in different operating conditions in healthy state and thus output health feature images.The healthy indicator was constructed using the differences between real and healthy images.The genetic algorithm was used to optimize the weights of multiple bearings in order to reduce the volatility of the comprehensive degradation curve in the healthy zone.The proposed method was tested on the unit's shaft system data and its validity has been proved.

关 键 词：水电机组 劣化评估 瞬时轨迹特征 条件对抗生成网络 遗传算法 

分 类 号：TV738[水利工程—水利水电工程]