Study Of Prediction Method For Ammonia Flue Gas Desulfurization Efficiency Based On Multi Parameters

With the development of national economy and improvement of living standards，energy demand increases increasingly, the release of SC^by the combustion of fossilfuels impacting on the environment has become more and more seirous. Flue gasdesulfuirzation is one of the important methods to solve the contradiction between airpollution and economic development in our country at present. Therefore, to select aset of advanced technology, production, economic reasonable, high efficiency andadvantageous to circular economy development and resource recycling flue gasdesulfuirzation process is very important. Ammonia lfue gas desulfuirzationtechnology due to its high efficiency, can both dcsulifiirze and denitrate，and sideproduct is ammonium sulfate fertilizer with high value and other advantages, which iswidely used in chemical industry, more and more energy and the electric powerenterprises.In view of the research of ammonia lfue gas desulfurization technology, thefollowing works has been done:(l)Partial least squares regression model(PLS), BPneural network model optimized by particle swarm optimization algoirthm(PSO-BP),support vector machine model optimized by simulate annealing algoirthm(SA-SVM)and least squares support vector machine model optimized by geneticalgoirthm(GALS-SVM) has been established respectively to predict efficiency,which takes several main operating parameters as the input vairables, such as the lfuegas lfow rate, ammonia concentration? liquid-gas ratio, spray liquid density?absorption liquid concentration, inlet gas temperature, ammonia consumption, flowof circulating pump, the pH value of spray liquid, prewashing tower? concentrationtank and so on.(2)Select20sets of practical monitoring data of a certain power plantto test the newly built model’s prediction precision.(3) Select another set of data forGALS-SVM and change one parameter continuously for multiple prediction and observing the relationship between the parameter and desulfuirzation efficiency.(4)To select another group of training samples, predict the desulfuirzation eiffciencywith the four artiifcial intelligence computation model, then compares their predictionaccuracy.The results show that：(l)The above four kinds of artificial intelligencetechnology model can predict the ammonia lfue gas desulfuirzation efficiency withhigh forecasting accuracy which has no obvious difference.(2)GALS-SVM model canbe applied to the study of ammonia desulfuirzation process.To sum up, artificial intelligence technology model can be applied to the study ofammonia desulfuirzation process，which provides the reference for the ammonia fluegas desulfuirzation technology research and operation parameter optimization.