基于多数据融合的水电机组劣化趋势概率区间预测  

Prediction of Deterioration Trend Probability Interval of Hydropower Unit Based on Multi-data Fusion

作　　者：王淑青[1,2] 翟宇胜 胡文庆 盛世龙 刘东 WANG Shu-qing;ZHAI Yu-sheng;HU Wen-qing;SHENG Shi-long;LIU Dong(School of Electrical and Electronic Engineering,Hubei University of Technology,Wuhan 430068,China;Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System,Hubei University of Technology,Wuhan 430068,China;School of Power and Mechanical Engineering,Wuhan University,Wuhan 430072,China;College of Energy and Power Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450045,China)

机构地区：[1]湖北工业大学电气与电子工程学院,湖北武汉430068 [2]湖北工业大学太阳能高效利用及储能运行控制湖北省重点实验室,湖北武汉430068 [3]武汉大学动力与机械学院,湖北武汉430072 [4]华北水利水电大学能源与动力工程学院,河南郑州450045

出　　处：《水电能源科学》2025年第2期201-205,共5页Water Resources and Power

基　　金：国家自然科学基金项目(52309111)。

摘　　要：传统的基于单一测点的预测模型无法全面反映水电机组的健康状态,这导致难以实现机组劣化状态的准确评估。对此,提出了一种基于多测点数据融合与概率区间预测的水电机组劣化趋势预测模型。首先,选取机组不同测点在各工况下健康运行的数据构成数据集,采用期望最大化—高斯混合模型(EM-GMM)拟合机组健康运行状态下的各监测量的概率密度分布;然后,计算待估样本在给定机组健康状态分布下的负对数似然概率,以作为劣化度指标;其次,采用熵权法计算各测点劣化度指标的权重,通过加权得到综合劣化度指标;最后,为确保预测结果的可靠性,利用多目标遗传算法(MOGA)优化高斯过程回归(GPR)模型代替传统的点预测模型,并使用不同的预测模型进行对比和评估,证明本文提出的模型具有更高的预测精度。Since the traditional prediction model based on a single measurement point cannot fully reflect the health state of a hydropower unit,this leads to difficulties in realizing an accurate assessment of the unit's deterioration state.In this regard,this paper proposes a prediction model for the deterioration trend of a hydropower unit based on the fusion of multi-measurement point data and probabilistic interval prediction.Firstly,the data from different measurement points of the unit running healthily under various operating conditions are selected to form a dataset.The expectation-maximization-Gaussian mixture model(EM-GMM)model is used to fit the probability density distribution of each monitoring variables under the healthy running state of the unit.Then,the negative log-likelihood probability of the samples to be estimated under a given unit health state distribution is calculated to serve as an indicator of the degree of deterioration.Secondly,the entropy weighting method is used to calculate the weights of the deterioration indicators at each measurement point,and the integrated deterioration indicators are obtained by weighting.Finally,to ensure the reliability of the prediction results,the Gaussian process regression(GPR)model is optimized using multi-objective genetic algorithm(MOGA)instead of the traditional point prediction model,and the different prediction models are used for comparison and evaluation to prove that the proposed model has a higher prediction accuracy.

关 键 词：水电机组 多数据融合 EM-GMM健康模型 劣化度指标 熵权法 概率区间预测模型 

分 类 号：TV734.21[水利工程—水利水电工程]