Modeling And Generalized Predictive Control Of Low-temperature SCR Denitrification System For Coke Oven Flue Gas

NOx emitted from coking process is one of the main sources of atmospheric pollutants.The NOx in the atmosphere can cause serious harm to human health,the environment,and buildings.Therefore,the ultra-low emission transformation of the coking industry is imperative.Selective Catalytic Reduction(SCR)technology is currently a widely used denitrification technology.However,the coke oven lowtemperature SCR denitrification system is a complex controlled object with nonlinearity,large hysteresis,and strong disturbance,and existing PID control and neural network control are difficult to achieve satisfactory control results.Therefore,studying databased modeling methods for denitrification systems and applying advanced process control algorithms to the control of denitrification systems is of great significance.This thesis takes the low-temperature SCR denitrification system of a coking plant as the research object,establishes a Single Input Single Output(SISO)model with ammonia flow rate as input and outlet NOx concentration as output,and applies the Step Generalized Predictive Control(SGPC)algorithm to the control of outlet NOx concentration.Firstly,the actual operation data of the denitration system is obtained through the cloud intelligent monitoring platform of the coking plant,and the original data is preprocessed by data screening,singular value detection and correction,data filtering and denoising.Aiming at the strong disturbance caused by reversing during coke oven operation,a modeling method of coke oven low-temperature SCR denitration system with intelligent supervisory level is proposed in this thesis.This method uses the Controlled Auto-Regressive Integral Moving Average(CARIMA)model to describe the controlled object,and combines Akaike’s Information Criterion(AIC)and Batch Least Squares(BLS)method to determine the delay and order of the model.By using the Forgetting Factor Recursive Least Squares(FFRLS)method to identify model parameters,an intelligent supervision level is introduced into the identification process to ensure the convergence of parameters and comply with the mechanism.Cluster analysis is used to analyze the identification results to determine the final model parameters.The cross validation of on-site collected data indicates the effectiveness of the model.The application of SGPC algorithm in coke oven low temperature SCR denitration system is studied.In this thesis,a SGPC constraint problem solving method based on Firefly algorithm(FA)is proposed.The optimization parameters of FA iteration are dynamically adjusted according to the brightness difference,which improves the convergence speed and optimization accuracy of the algorithm.Combined with the data collected on site,the influence of SGPC controller parameters on the control performance trend is analyzed,and the recommended range of controller parameters suitable for coke oven low-temperature SCR denitration system is given.In the simulation environment,the SGPC controller and PID controller are compared under the condition of model mismatch and different disturbance.The results show that the SGPC controller has better robustness and anti-interference performance.Deploy the designed SGPC controller through edge servers to industrial sites for practical testing.In order to reduce the exceeding time of outlet NOx concentration,this thesis proposes a segmented control strategy,which increases the controller’s rapidity by adjusting the controller parameters when the outlet NOx concentration exceeds the limit.Aiming at the problem of drastic fluctuation of control quantity and output quantity caused by coke oven reversing,reversing filtering strategy is proposed in this thesis,which reduces the fluctuation degree of output quantity and improves the stability of control.The analysis of the actual control effect shows that,compared with the original neural network controller in the coking plant,under the control of the SGPC controller designed in this thesis,the NOx concentration at the outlet not only fluctuates less,but is closer to the set value and consumes less ammonia water,which effectively improves the automation level of the industrial site and reduces the denitrification cost of the enterprise.