| With China’s policy on environmental protection requirements of increase years,thermal power unit NOx in flue gas emissions had incorporated into strict regulation.selective catalytic reduction(SCR)of the flue gas technology because of its high rate,reliable technology,the advantages of simple structure and small ammonia escape rate has become the main choice of coal-fired power plant boiler control NOx emissions.SCR fluegas control system is an important part to ensure the safe and continuous operation of the system and meet the performance indexes of the system.The model of SCR system is the basis for the research of control strategy.In order to obtain an accurate and effective SCR system model,this paper adopts different means to study the modeling method of SCR system,and compares and analyzes the accuracy and generalization ability of the model.The main research contents include:(1)By analyzing the mechanism of SCR reaction,a SCR mechanism model was established.Using the operation data of SCR system of 350 MW unit and the adaptive particle swarm optimization algorithm was adopted to identify the model parameters and further analyze the model performance.(2)The data modeling method of SCR system based on least squares support vector machine was studied.Firstly,the main influencing factors of SCR system have analyzed by mutual information method,and the model input variables were further selected.Then,based on the selected input variable set of the model,the data model of SCR system was constructed by using the least square support vector machine,and the performance of the model was further analyzed.(3)Combined with the characteristics of the mechanism model and the data model,the data model is used to correct the deviation of the mechanism model,and the output of the least square support vector machine model is used as the error estimator and the output of the mechanism model is superposition as the output of the mixed model.The mixed model of SCR system was established by using this method,and the performance of the model was analyzed.The results show that the hybrid model combines the advantages of mechanism model and data model,and has good accuracy and generalization ability... |