Research On Pollutant Emission Characteristics In Power Plant Boiler Based On Reverse Modeling

Affected by the amount of energy resources, coals occupy a large proportion in the composition of China’s primary energy. Combustion is the main way of using coals. In the coal combustion process, sulfur oxides, nitrogen oxides, carbon monoxide, organic pollutants and other pollutants will be released. Electricity supply is based on thermal power. Since coal consumption of the power plant coal-fired boilers is large, they are the focus of the pollutant emission control objects. However, thermal power with nonlinear, large inertia and other characteristics is typical complex thermodynamic system, the traditional modeling methods have significant limitations in the modeling of complex thermal system. The inverse modeling method is used to predict the emissions concentration of nitrogen oxides and dust.Factors affecting NOx emission characteristics were analyzed in detail. Combined with DCS system online monitoring point parameters, input variables are chosen to establish the model of NOx emission characteristics. The method of kernel principal component analysis (KPCA) is introduced to extract independent variables and it makes up for the inadequacies of the principal component analysis (PCA) method in large amount computation of eigenvalues and eigenvectors and dealing with nonlinear problem. The projection pursuit regression model(PPR) based on the Hermite polynomials are used to predict the concentration of nitrogen oxide emissions. The results show KPCA-PPR model has a higher accuracy.As the importance of the samples, the Projection Pursuit method(PP) is used to extract the characteristic variables and the interference of man-made factors is avoided. The results provide the basis for determining factors of the dust concentration. PP achieves a comprehensive evaluation of the samples, which make the model more stability. Since the advantages of artificial neural networks in modeling accuracy and generalization ability, dust emission concentration prediction model is built, which provide new ideas and approaches for modeling.On the basis of the above research results, actual readable online operating parameters are taken as the initial the input variables.The various operation parameters which are extracted by the method of PP and KPCA are used as the input variables. The concentration of nitrogen oxides and dust are used as output variables of the comprehensive model. Tt can be seen that the model is simple, strong applicability, and can meet the requirement of accuracy.