Study On Integrated Control Method Of NO_x And SO₂ In Coal-fired Flue Gas Based On Neural Network

At present,coal-fired units achieve ultra-low emission standards by upgrading to a single technology,but this single technology upgrade leads to increased energy consumption of the entire pollutant removal system,increased plant operation and maintenance costs,and reduced overall economics of the system.Therefore,an important issue for coal-fired power stations is how to control flue gas pollutants in a more energy-efficient and efficient manner.With the increasingly strict environmental protection standards,coal-fired units are bound to adopt integrated pollutant control technology.In this paper,with the support of the National Key Research and Development Program—“Research and Engineering Demonstration on Integrated Coal-fired Boiler Pollutants（SO₂、NO_X、PM）Control Technology”,models of the SCR denitration and the absorption tower system were established based on the operating data by combining the denitration and desulfurization process mechanism analysis.On this basis,the genetic algorithm is used to optimize the operating parameters,and the system’s operating cost is the lowest when the emission standards are met,providing an operating strategy for this integrated control.The main contents include:（1）Stability factor is introduced as the criterion for steady-state operating conditions.Unit load and flue gas volume are selected as judging parameters.The actual operating data of the unit under steady-state operating conditions is collected and the data is pre-processed.（2）Analysis of pollutant removal mechanism and influencing factors of SCR and absorption tower,calculation of Pearson correlation coefficient of influencing factor and output value was completed.Wherein the absolute value of the Pearson correlation coefficient injection amount of ammonia,the temperature of the flue gas,liquid-gas ratio,the depth of immersion,the ozone flow rate of greater than 0.5;The input parameters of the model are selected through a combination of mechanism analysis and calculation results of Pearson correlation coefficient;BP neural network and support vector machine were used for modeling,and the accuracy of the model was compared and analyzed.The results show that the BP neural network model has better prediction effect.Among them,the RMSE of the SCR model is 4.43 mg/m³,the MAPE is 5.26%,the RMSE of the absorption tower model is 4.27 mg/m³,and the MAPE is 8.39%.（3）Through the comparison of model accuracy,the genetic algorithm is used to optimize the BP neural network model with better accuracy,which further improves the model accuracy.The RMSE and MAPE of the SCR model are reduced to 2.73 mg/m³ and3.91%respectively,and the RMSE and MAPE of the absorption tower model are reduced to 2.17 mg/m³ and 6.64%;By introducing the reduced concentration,an operating cost calculation model for the integrated control route was established,and the ton pollutant removal cost was taken as the objective function,the operating parameters and various operating cost were calculated and optimized by genetic algorithm when the ton pollutant removal cost is the lowest By comparing and analyzing the two operating modes of SCR independent control and integrated control to achieve the operating cost under the set environmental protection standard,the analysis results show that integrated control has obvious advantages in control effect and economy.（4）Based on the optimized BP neural network model,the optimization guidance software is developed using the Py Qt library.Through the simple GUI user interface,the optimized calculation results of integrated pollutant control are displayed in the form of images and data,which is convenient for practical application.