Modeling Of SO₂ Emissions From Wet Desulfurization Systems For Coal-fired Unit

Coal has a dominant position in the national energy structure,but coal burning also produces a large amount of SO₂ gas,if directly emitted into the air will cause serious pollution to the atmospheric environment,so the flue gas produced by coal must be desulfurized.At present,the flue gas desulfurization method mainly used in the coal-fired unit is limestone-gypsum wet desulfurization,which has the advantages of high desulfurization efficiency,safety and reliability,low cost,etc.,but because the desulfurization system is greatly affected by the load fluctuation of the unit,the chemical reaction in the absorption tower is also more complicated,which brings certain difficulties to the operator to control the stable operation of the wet desulfurization system.Aiming at the problem that it is difficult to achieve stable control in the wet desulfurization system,this paper uses the data-driven method to establish a prediction model for the SO₂ concentration at the outlet of the absorption tower of the desulfurization system.By introducing the working principle of the wet desulfurization system for coal-fired units,the initial auxiliary variables related to the so2 concentration at the outlet were selected.The data preprocessing of auxiliary variables and the delay compensation of each input variable using mutual information algorithms finally determine the input of the model.The regularization limit learning machine is used to establish a prediction model,and the ephemeral algorithm is used to optimize the model parameters.The comparative prediction model is established by using the limit learning machine,the long short-term memory network and the least squares support vector machine,and the operation data of a power plant in Changzhi,Shanxi Province is used for simulation experiments,and the results show that the regularized limit learning machine has the advantages of high prediction accuracy,strong generalization ability,fast training speed and easy to achieve compared with other algorithms.The Stacking integrated algorithm is used to further improve the model accuracy,and multiple regularized limit learning machines are trained from the data of different time periods from the historical data of the power plant as the base model of the integrated algorithm,and the output of each base model is used as the input of the second layer model and then predicted,and the second layer model adopts the multivariate linear regression algorithm.Simulation results show that the Stacking integration model combines the advantages of multiple base models,so that the prediction error of the integrated model on the test set is lower than that of any base model.The prediction model established by the integrated algorithm can accurately predict the change of export SO₂ concentration,which provides a new idea and method for data-driven modeling to be applied to desulfurization prediction.