Prediction Of Petrochemical Waste Gas Emissions Based On Machine Learning

The atmospheric environment is one of the basic elements of human survival.The quality of atmospheric environment not only affects the ecosystem,but also concerns the sustainable development of social economy.Petrochemical enterprises are industries with relatively serious air pollution.The effective prediction of pollutant emissions from petrochemical waste gas can provide a basis for early warning and reduction measures by relevant assessment departments.The machine learning theory based on artificial neural network and support vector regression is applied to predict pollutant emissions from petrochemical waste gas for the first time in this paper.The paper relies on the online monitoring data of exhaust gas emissions of a petrochemical enterprise to predict the daily maximum concentration of SO₂ and NO_x in petrochemical waste gas.Firstly,the BP neural network is used as the model for predicting the concentration of petrochemical waste gas.Based on a large number of experimental analyses,the network structure is determined,and the characteristics of the model are analyzed according to three kinds of evaluating standards,which are Root Mean Squared Error,Normalized Mean Error and Coefficient of Determination.The additional momentum method,the learning rate decreasing method and the LM algorithm are used to optimize the petrochemical exhaust emission prediction model based on BP neural network to improve training speed and accuracy.Secondly,based on the support vector regression theory,a model of petrochemical exhaust gas pollution emission prediction based on support vector regression is established.Finally,the processing results of support vector regression are compared with the prediction results of the BP neural network.The results show that BP neural network and support vector regression can describe the complex nonlinear relationship of petrochemical exhaust gas pollutants and achieve prediction results.Support vector regression has higher predictive ability than neural network when dealing with small sample and high-dimensional problems in petrochemical emission forecasting,and both BP neural network based on LM algorithm and SVR both can meet the prediction accuracy of engineering requirements.The research in this dissertation proves the feasibility of using artificial neural network and support vector regression to predict the emission of petrochemical waste gas,which provides a new idea and method for the early warning work of gas pollution in petrochemical industry,and then improves the efficiency.