Modeling And Optimal Control Of SCR Flue Gas Denitration System

With the increasing problems of air pollution represented by acid rain and haze,environmental pollution control has gradually become an important aspect of energy production.In coal-fired power plants of China,SCR denitrification technique is one of the major means to control NOX pollution for its maturity and extensive use.In the deep study of SCR denitrification technique,the method of experimental modeling,which concentrates on the characteristics of SCR denitrification system under variable operating conditions,should be adopted.To establish effective SCR denitrification system model,input and output data that can reflect the dynamic characteristics of the system are chosen.The model is established as the basis of optimal control process,which is an important way to improve denitrification efficiency and of great guiding significance in designing and operating SCR denitrification equipment.The dissertation takes variable load operation normal of thermal power plant as background and reviews SCR denitrification technique and its applications.Based on the consideration of the model's strong generalization ability,broad popularization performance,fast computation speed,and strong applicability of online work,nonlinear model is built through LSSVM.For the multicollinearity of variables,optimal approach of similarity function is firstly adopted to remove variable redundancy data;then,to better reflect the dynamic characteristics of system objects,kernel principal component analysis and multivariable process are combined to monitor the dominant factor of NOX concentration at SCR reactor inlet.Parameters in the model are determined by grid search combined with particle swarm optimization(PSO)algorithm,LSSVM prediction model is built.For unit variable load requirements,the dissertation predefines updated threshold that can adapt changes according to boiler load variation,combining sliding window method.On the basis of the oldest data removing principle and forecast error for adaptive update model parameters,two dynamic models are established.Above modeling approaches are respectively applied and tested in SCR denitrification system modeling.The results indicate that dynamic model has higher forecast accuracy and strong generalization ability.Finally,the model is applied in predictive control algorithm and transformed into nonlinear optimization problem.PSO algorithm is used in optimizing spray ammonia flow to make NOX concentration at SCR reactor inlet track the set value accurately,and to improve denitrification efficiency.The experimental results show that compared with traditional PID control mode,this method is able to achieves better control effect.