基于稀疏高斯过程回归健康模型的抽蓄机组轴系健康状态评估方法  

Health State Assessment Method for Shaft System of Pumped Storage Unit Based on Sparse Gaussian Process Regression Health Model

作　　者：张启航 张孝远[1] 张宇翔 高玉峰 马驰 ZHANG Qi-hang;ZHANG Xiao-yuan;ZHANG Yu-xiang;GAO Yu-feng;MA Chi(Colege of Electrical Engineering,Henan University of Technology,Zhengzhou 450001,China)

机构地区：[1]河南工业大学电气工程学院,河南郑州450001

出　　处：《水电能源科学》2025年第3期173-176,共4页Water Resources and Power

基　　金：国家自然科学基金项目(52379088);河南省自然科学基金项目(232300421207);河南工业大学自科创新基金计划项目(2022ZKCJ04)。

摘　　要：抽水蓄能机组(以下简称抽蓄机组)在高比例新能源电力系统中担任调能角色,对其设备开展实时健康评估对机组本身及其互联电力系统的安全均有重要意义。针对当前研究在揭示机组性能退化的不确定性、状态评估实时性方面存在的不足,提出了一种结合小波阈值降噪(WNR)和稀疏高斯过程回归(SGPR)的抽蓄机组健康状态评估方法。该方法首先采用WNR对监测数据进行降噪以提升数据质量,然后采用抽蓄机组健康运行时刻的数据基于SGPR构造抽蓄机组的健康基准模型(HBM)。在评估时刻,采用在线采集的性能参数与HBM预测得到的健康性能参数的偏差来量化机组的劣化情况。区别于传统的点估计方法,SGPR的输出作为机组性能的合理区间,可量化机组劣化中的不确定性。实例验证表明,与其他方法相比,所提方法在95%置信水平下,拥有最好的区间覆盖率及狭窄的区间宽度,并在计算耗时上相较于传统的高斯过程回归(GPR)方法降低了90%。工程实际检修数据验证了所提方法的有效性。Pumped storage units play the role of energy regulation in a high proportion of new energy power system,so it is of great significance to carry out real-time health assessment of their equipment for the safety of the units themselves and the interconnected power systems.Aiming at the shortcomings of current research in revealing the uncertainty of unit performance degradation and the real-time state assessment,a method combining wavelet threshold noise reduction(WNR)and sparse Gaussian process regression(SGPR)is proposed to evaluate the health state of pumped storage units.Firstly,WNR is used to reduce the noise of the monitoring data for improving the data quality.And then the health benchmark model(HBM)of the pumped storage unit based SGPR is constructed based on the data of the healthy running time of the pumped storage unit.At the evaluation stage,the deviation between the performance parameters collected online and the health performance parameters predicted by HBM is used to quantify the deterioration of the unit.Different from the traditional point estimation methods,the output of SGPR can be used as a reasonable interval of unit performance to quantify the uncertainty of unit deterioration.The experimental results show that the proposed method has the best interval coverage and narrow interval width at 95%confidence level,and the calculation time is reduced by 90%compared with the traditional Gaussian process regression(GPR)method.Finally,the effectiveness of the proposed method is verified by the actual project maintenance data.

关 键 词：健康状态评估 稀疏高斯过程回归 小波降噪 抽水蓄能机组 

分 类 号：TV734.1[水利工程—水利水电工程]