Research On Modeling Of Desulfurization System In Coal-fired CFB Boiler Based On Massive Data

Evolved gases such as SO₂ and NO_x during coal combustion cause pollution to the environment.How to reduce such pollutants emitted by coal-fired power plants has become a hot research topic.In boilers,sulfur in coal mainly exists in the form of SO₂after combustion.Accurate monitoring and control of SO₂ emission is the key work in the actual operation of coal-fired power plants.Based on the nonlinear and strongly coupled operating parameters of the DCS control system of a coal-fired power plant,a CFB boiler with 150t/h steam in a thermal power plant is selected as the research object,and the economic evaluation model and furnace outlet of the coal-fired boiler’s desulfurization system are established respectively.The main research contents of the SO₂ concentration prediction model are as follows:On the basis of the desulfurization operation specification diagram of the power plant,statistical calculations were carried out for the two types of desulfurization costs in the furnace and outside the furnace of 1#boiler,and the combined desulfurization cost was obtained.The optimization of the two angles provides certain optimization suggestions for the desulfurization operation of thermal power plants,and also provides a basis for the prediction model.In order to screen the input parameters of the SO₂ prediction model,the boiler parameters recorded by the DCS system of the power plant were explored.The DCS system includes boiler evaporation,bed temperature,primary and secondary air volume and air temperature,and other important operating parameters that affect the generation of sulfur.The relevant influencing factors of SO₂ emissions are screened out from the DCS system based on chemical reaction mechanism.After that,these factors are classified into inlet state parameters,inlet reaction parameters,reaction condition parameters and outlet state parameters according to the different position and influence mechanism.Since there are 30 groups of input parameters,the principal component analysis method is used to simplify the data and reduce the dimensionality,and the input parameters are converted into 6 groups of principal components through this analysis method.The simplified principal component is selected as the input parameter,and the best BP neural network model is sought by comparing the prediction errors of different structures.Finally,it is determined to use the best predictive structure with the relu function,500 training epochs,the single layer,8 nodes,and 0.005 learning rate.Under this structure,the test and verification data show that when boiler evaporation,primary air volume,flue gas oxygen content,etc.are used as input parameters,the model with different parameter combinations can achieve a prediction relative error of 2.7%,which can be better Forecast the changing trend of pollutant emissions.