基于GMM与NSET优化算法的设备参数预警研究  

Research on Equipment Parameter Early Warning Based on GMM and NSET Optimization Algorithm

作　　者：袁雪峰 马成龙 陈世和 YUAN Xue-feng;MA Cheng-long;CHEN Shi-he(Smart Power Generation Research Center,China Resources Electric Power Technology Research Institute Co.,Ltd.,Smart Power Generation Research Center,Shenzhen 518001,China)

机构地区：[1]华润电力技术研究院有限公司智慧发电研究中心,广东深圳518001

出　　处：《控制工程》2022年第6期1058-1064,共7页Control Engineering of China

摘　　要：针对工业流程存在难以监测非重要数据和设备劣化状态的问题,提出了一种基于高斯混合模型(GMM)和非线性状态估计(NSET)优化算法的设备参数预警方法,用于在设备运行过程提前发现设备的故障趋势。首先,根据设备特点对历史运行数据进行预处理;然后,利用高斯混合模型聚类获得聚类中心;最后,利用非线性状态估计建模,将聚类中心与抽样样本共同作为记忆矩阵,对实时运行数据进行预测和偏差分析,对偏差超过经验阈值的传感器报警。以某1000 MW火电机组引风机为例,选取历史典型运行状态数据进行分析,分析结果表明,该方法能有效实现传感器参数预警。Aiming at the problem that it is difficult to monitor non-important data and deteriorating state in industrial processes,an equipment parameter early warning method based on Gaussian mixture model(GMM)and nonlinear state estimation(NSET)optimization algorithm is proposed.The purpose is to detect the deterioration of the equipment in advance during the operation of the equipment.Firstly,the historical operation data are preprocessed according to the characteristics of the equipment.Then,Gaussian mixture model is used to obtain the clustering center.Finally,the nonlinear state estimation is used to establish the model.The clustering center and sampling samples were used as memory matrix to predict and analyze the deviation of real-time running data.If the deviation exceeds the empirical threshold,the sensor will be alerted.The induced draft fan of a 1000MW thermal power unit is taken as an example.Historical typical operating status data are selected for analysis.The analysis results show that this method can realize sensor parameter early warning effectively.

关 键 词：工业设备 高斯混合模型 非线性状态估计优化 传感器参数预警 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]