Research On Modeling Method Of Denitration System Based On SOA-IRBF Neural Network

At present,the state has formulated more and more strict NOx emission standards for coal-fired power plants,and strengthened supervision and punishment.In order to meet the national emission standard,the reasonable regulation of SNCR process urea solution in SNCR-SCR denitration system and reduce economic cost,it is necessary to accurately predict the NOx concentration at the inlet of SCR.Therefore,this paper proposes a denitration system modeling method based on the combination of seagull algorithm(SOA)and improved RBF neural network(IRBFNN)to accurately predict the NOx concentration at the inlet of SCR.The modeling algorithm uses mutual information method to calculate the correlation between 13 auxiliary variables and SCR inlet NOx concentratio n and arrange them in descending order.The forward search strategy is adopted to eliminate irrelevant variables and redundant variables and screen out 5 auxiliary variables,so as to avoid too few or too many variables that can not fully express the inter nal structure or over fitting of the system,thus affecting the prediction accuracy of the model;The combination of delay correlation function and greedy search is used to estimate the delay of the selected auxiliary variables to avoid the reduction of mo del accuracy caused by delay problem;The auxiliary variable of time reconstruction is used as the input of the network model with different hidden layers,and the nitrogen oxide concentration at the inlet of SCR is used as the output of the model to establish the prediction model.Three different optimization algorithms are used to optimize the structural parameters of the model and calculate the evaluation standard values of the network models with different hidden layers.Through the comparison of images and standard values,it can be verified that the prediction accuracy of the RBF neural network model with double hidden layers is the highest,The results show that the modeling method of denitration system based on SOA-IRBF neural network not only improves the prediction accuracy of the model,but also speeds up the running time of the model.