Prediction of SCR Catalysts Performance in Coal-Fired Power Plants Based on Fusion of Multi-source Information

Abstract

In this paper, an integrated algorithm was proposed to evaluate the DeNOx ability of SCR reactor using the operational data of power plant, including the identification of steady state based on R-statistic, the detection of the wrong points and the classification of working conditions by CNN-clustering algorithm. Then the performance profiles of the SCR reactor were extracted from the mass and mixed raw data. Furthermore, the catalyst samples after 0 days, 540 days and 1246 days of operation were collected from the first catalyst layer, and the degradation of the catalyst activity and the micro geometrical characteristic were obtained by the laboratory test. Finally, the FCM fuzzy clustering algorithm was used to fuse the health characteristics of the SCR catalyst layer from multi-source information, such as DeNOx ability of the SCR reactor, catalyst activity, the service of catalyst, and the health evaluation system of the catalyst layer in a full lifecycle was obtained.