Desulfurization Efficiency Prediction Based On Clustering And Partial Least Squares Support Vector Machine

Since the 21 st century,with the rapid development of China's economy,the demand for energy and electricity has also risen.At present,coal-fired thermal power generation is still the mainstay of market supply,and harmful gases such as sulfur dioxide and nitrogen dioxide produced by coal combustion not only cause atmospheric pollution,It also seriously harms human health.For this reason,people have studied and implemented various desulfurization schemes,and all have achieved good desulfurization results.The desulfurization efficiency can reflect the performance of the system very well.At present,power plants mainly use testing equipment to measure flue gas.Considering equipment damage and high maintenance costs,soft measurement technology for desulfurization efficiency has been proposed and plays an increasingly important role.Aiming at the limestone-gypsum wet desulfurization technology,this dissertation proposes a partial least squares support vector machine based on cluster analysis to predict the desulfurization efficiency.The historical data in the DCS database system of a power plant in Shanxi was selected to simply screen the factors affecting the desulfurization efficiency,and the model was built and trained through the MATLAB platform.The cluster analysis,partial least squares(PLS),and least squares support vector machine(LSSVM)are combined and compared with a single PSO-LSSVM prediction model.The research shows that the mean square error of the proposed models is between 0.01 ~ 0.02,and the latter reaches 0.02 ~ 0.04,which also strongly illustrates the accuracy and validity of the research model in this dissertation,and the possibility of further application in engineering,that also laid the foundation for further model optimization and modification and upgrade.