Research on abnormal search method of monitoring parameters for power plant equipment based on cluster analysis

Abstract

In order to maintain the security, reliability and economy in actual production process and evaluate the health of operation status of power plant equipment, an abnormal search method of online monitoring parameters is proposed in this paper. By containing an integration of a large number of historical data and real-time data, the abnormal search method of online monitoring parameters can search out some special data or data segments that are significant and different from other data in the time subsequence of the monitoring parameters. In the abnormal parameter search process, it needs a segmentation of the time sequence based on some important points and the extraction of eigenvalues with the sub-patterns of the time subsequence. Then it is necessary to map the time subsequence to a high-dimensional feature space and grub the anomaly information from the perspective of model so as to obtain the set of anomaly parameters. It is of great significance to search out the obviously inconsistent data or behaviors from the other general patterns in the time series of online monitoring parameters for the early failure warning of the large and complex equipment.