Study On Furnace Coal Quality Parameters And No_x Emission Prediction Under Coal Blending Combustion

This paper is dedicated to the study of the whole process of coal blending,constructing a power blending model before coal blending and a fast calculation method of furnace coal quality,and studying the NO_x emission characteristics of coal blending after combustion.In order to adapt to the complex and variable coal combustion in power plants and the lack of guidance on coal blending scheme,a targeted coal blending model is constructed.Firstly,k-means cluster analysis and normal distribution fitting are performed for coal quality parameters such as incoming coal heat,volatile fraction and sulfur to obtain coal quality distribution functions and interval probabilities;secondly,t-hypothesis test method and linear fitting are used to analyze the errors between test and theoretical weighted values of 30 groups of blended coal quality to obtain a blended coal quality calculation method with high applicability;the constraint boundaries and weights of coal quality indexes for boilers are determined,and the The blending model with coal blending target as input can output the coal blending guiding scheme,including coal supply type and blending ratio,and the multi-objective fuzzy evaluation method is applied to the evaluation of the blending scheme.In order to better solve the problem of lagging time of furnace coal sampling and testing,the research idea of forward tracking and reverse traceability is proposed.The correlation analysis between the incoming coal quality and the furnace coal quality testing results under blending conditions is carried out based on the tracking and traceability from the incoming coal to the combustion in the furnace.The results show that the coal quality parameters obtained from the fast calculation model of furnace coal are more comprehensive and the errors are more in line with the requirements of industrial applications.To establish a prediction model of coal combustion pollutants represented by NO_x,the factors affecting NO_x generation are screened and pre-treated.Based on the NO_x generation mechanism,101 influencing variables related to furnace air distribution,boiler load and flue gas oxygen content were screened,and Correlation Analysis(CA)and Principal Component Analysis(PCA)were performed to reduce the dimensionality,select the key variables and reduce the data information of variables coupling.The results showed that 59 variables had correlation coefficients no less than 0.40 with NO_x,and the first 8 principal components containing 98.34%of the original data were selected as input variables for the NO_x concentration prediction model.In this paper,BP neural network models based on correlation analysis and principal component analysis were established with 7489 sets of samples with time interval of 1 min to predict and analyze the inlet NO_x concentration and single component NO concentration of SCR system,respectively.The average relative error of the CA-BP model was 2.22%and the relative error of the single-component NO concentration was 1.54%,while the relative errors of the PCA-BP model were 2.46%and 1.76%,respectively.The genetic algorithm(GA)was used to optimize the model parameters,and the relative error of the optimized PCA-BP model for NO_xconcentration prediction was 2.15%,which indicates the higher prediction accuracy of the optimized model.