The Modeling And Application Of Data-based Process NO_x Prediction

China's energy structure is coal based,leading to the main power production is still fire power generation.However,there is an important problem of relying on the power plant to generate electricity is environmental pollution,such as NO_x,SO_x and other pollutants,which can endanger the ecological environment and human health.The latest revision of the atmospheric?Pollutant Emission Standard in Thermal Power Plants??gb13223-2011?is based on the Models-3/CMAQ of the integrated air quality model system to predict and evaluate the environmental impact.Since July 1,2014,the key areas,coal-fired boiler smoke emission of NO_x limit value of 100mg/Nm3,which puts forward higher requirements for energy saving and emission reduction of coal-fired power generation.At present,the domestic and foreign research for controlling NO_x emissions can be divided into three aspects:the improvement of boiler combustion technology;ammonia injection into furnace without catalyst;ammonia water injection system with catalyst.This article belongs to the study of ammonia spraying system,which is the second or third aspect.the most widely used denitration technology in industry is selective catalytic reduction?SCR?,the NO_x in the flue gas and ammonia chemical reaction in the role of the catalyst,leading to NO_x decomposition into water vapor and nitrogen molecules?N₂?,Both of them are inert gases in the atmosphere,in order to achieve the purpose of protecting the environment.However,it is difficult to obtain accurate and rapid parameters under the limitations of existing funds and technologies,and soft sensor technology can solve this problem well,that is to choose some easily measured variables and to construct a mathematical relationship to predict and estimate them.Modeling based on process mechanism and based on process data are two kinds of soft sensor modeling,because of the complex mechanism of the power plant,it is necessary to establish the soft measurement model based on the process data,which is to extract the useful information from the input data and construct the mathematical relation between the auxiliary variable and the dependent variable.Based on the study of traditional statistical regression method and the method based on artificial intelligence machine learning,this paper puts forward a new soft measurement forecasting model,a hybrid SPSS-PSO-SVM model to predict the NO_x emission of power plant,and the introduction of statistical software SPSS can intuitively observe and analyze the correlation between variables.The principal component matrix is calculated by factor analysis to reduce the dimension of input variables and eliminate the correlation.Taking the data of a 600MW power plant boiler as the running object,the main components are extracted by SPSS in order to eliminate the correlation between variables and to remove the variables that affect them lightly,so as to avoid the deviation of the conclusions;PSO is optimized to obtain the optimal parameter combination in order to achieve the optimal prediction accuracy.The MATLAB simulation of SVM is designed to predict the NO_x content of boiler exit at present,then making the amount of ammonia injection.