Study On Soft Sensor Prediction Of Flue-gas Emission Based On Process Monitoring

Mankind is faced with increasingly serious environmental problems in which air pollution is an important aspect. The emission of exhaust gas and dust from thermal power plant is an important source of atmospheric pollution. In order to limit the emission of pollutants from thermal power plant, the country has developed stringent pollutant emission standard. Sulfur dioxide is the main gaseous pollutant from thermal power plant. The study of this subject is the sulfur dioxide emission of the thermal power plant wet limestone-gypsum flue-gas desulfuration system. This paper, based on a comprehensive understanding of thermal power plant desulfurization system,analyzes the main factors affecting the desulfurization efficiency and the concentration emission of sulfur dioxide.Currently, in the thermal power plant, the mainly monitoring equipment for gas pollutants is the flue-gas continuous emission monitoring systems. The flue-gas continuous emission monitoring systems reflect the emission data and report it in time to thermal power plant and regulatory authorities. The emission concentration of flue gas pollutant, obtained by the flue gas continuous emission monitoring system, is measured through a dedicated analyzer. Based on relevant operating parameters from desulfurization system, the article creates a model to predict the desulfurization efficiency and the emission concentration of sulfur dioxide by making use of soft sensor technology.The stage of soft sensor modeling is the focus of this article. This paper, by means of the analysis of the factors affecting the desulfurization efficiency and based on the collected data and the actual situation, selects the slurry pH value, the emission concentration of sulfur dioxide in the desulfurization tower entrance, the temperature of flue gas in the desulfurization tower entrance and other eight parameters as the inputs of the soft sensor modeling. This paper selects two methods: BP neural network as well as support vector machine, and build the desulfurization efficiency prediction model by both of them. The result shows that both methods could achieve a certain predictive effect, but the support vector machine after parameter optimization has the best predictive performance. The parameter optimization result is that penalty parameter is 0.75786 and kernel parameter is 4.5948. The mean square error andaverage relative error of the optimizing support vector machine model predictions are0.179 and 0.367%. Finally, through OPC(OLE for Process Control) technology, this paper implements data exchange between MATLAB and configuration software of the pollution process monitoring system, in order to bring the advantages of two software-MATLAB and Kingview into full play. Theoretically, this paper realizes the online prediction of the desulfurization efficiency and the emission concentration of sulfur dioxide.