Prediction And Optimization Of SO₂ Emissions From Double Alkali Desulfurization Tower Based On RBF Neural Network

This subject “ prediction and optimization of SO2 emissions from double alkali desulfurization tower based on RBF neural network ” comes from a project item of Jiangsu Sanfangxiang Group —exploitation of emission concentration prediction and control system for flue gas desulfurization tower of coal-fired boiler.Subject to the enterprise is currently running a set of 6 # desulfurization tower 23 t / h boiler double alkali desulfurization system as the object of study,fully investigate the situation at the scene,and the relevant results in a timely manner to record.According to the historical operation data of desulphurization system and the experience of desulfurization tower emission concentration prediction,analyze the relationship between SO2 emissions from double alkali desulfurization tower and the operating parameters of the desulfurization system.First of all,through the study of the concrete project of a double alkali desulfurization system in Sanfangxiang Group,the basic flow and reaction mechanism of the double alkali desulfurization process were analyzed,and in the field conditions the main operating parameters of SO2 emissions in the double alkali desulfurization tower were pointed out.And the influence of these main operating parameters on SO2 emissions was analyzed.On the basis of this,the method of artificial neural network was used to study the SO2 emissions of the double alkali desulfurization tower.In this paper,mainly through the study of the theoretical basis of artificial neural network,the principle of RBF(Radial Basis Function)neural network algorithm is analyzed.The RBF neural network adopts self-organization selection center method,and variable gradient BP(Back Propagation)neural network using the traditional F-R correction and BP neural network with LM(Levenberg-Marquardt)algorithm optimization.The 100 sets of valid data were collected as the neural network training samples,and 15 sets of data were selected as the test samples.The SO2 emissions of the double alkali desulfurization tower was predicted on the MATLAB 2014 a platform.The results show that the RBF neural network has a good advantage over the other two networks in terms of prediction accuracy,training speed and generalization ability.Therefore,the RBF neural network is used to forecast and optimize the SO2 emissions from desulfurization tower smoke export.Then,an RBF neural network model for predicting and optimizing SO2 emissions of double alkali desulfurization tower was established.The 11 main desulphurization parameters of the desulfurization process were used as the input weight of the network.The influence of the conversion of SO2 concentration on the flue gas outlet in the RBF neural network model with the change of the desulfurization parameters was analyzed in detail.In the optimization prediction of the SO2 emissions of the desulfurization tower,the SO2 emissions from the flue gas outlet of the desulfurization tower is mainly considered as the influence of the change of the flue gas temperature,the p H value and the liquid / gas ratio.The results of optimization prediction show that RBF artificial neural network can not only predict SO2 emissions from flue gas outlet of desulfurization tower,but also can optimize the SO2 emissions from flue gas outlet of desulfurization tower to find the optimal desulfurization scheme.Finally,a set of SO2 emissions forecasting system of double alkali desulfurization tower based on RBF neural network is designed by MATLAB software platform,and a convenient and simple GUI user interface is developed.People only need to enter the operating parameters in the window,the forecast results will be displayed in the form of data.The model is simple and easy to be applied in practical engineering.