Dynamic Modeling of Nox Emission from Power Plants Considering Input Variable Time-Delay and Feature Extraction

Abstract

Aiming at the problem that it is difficult to effectively predict the concentration of nitrogen oxide emissions from boilers under the changing load conditions of power plants, this paper proposes a dynamic modeling method of nitrogen oxide (NOx) emissions from power plants, which considers the delay-feature extraction of input variables. This method combines principal component analysis and Gaussian regression modeling, and uses principal component analysis to extract the feature information extracted from the input variable data, and the current value and historical sequence value of extracted information are used as the input of Gaussian regression model. In addition, the historical time series value of boiler NOx concentration is added to the input of the model as feedback data. Taking the combustion system of a 1,000MW ultra-supercritical unit boiler as an object, and combining with the actual operation data on site, a dynamic model of boiler NOx emission is established. The experimental results show that the built boiler NOx emission dynamic model has high prediction accuracy and strong generalization performance, which has a certain reference value for boiler NOx emission modeling and intelligent control research.