Coal blending optimization model for reducing pollutant emission costs based on Support Vector Machine

Abstract

Factors such as the pollutant formation, pollution emission punishment and pollutant control devices are considered to optimize the coal blending method for a 300 MW boiler unit. The support vector machine (SVM) is used to establish the pollutant formation prediction model for the coal–fired boiler. Moreover, the model built above is trained and verified based on the actual operation data. Then the genetic algorithm is applied to optimize the coal blending method with the coal price to achieve the lowest operation cost. It can be concluded from the results that the precision of the prediction model is relatively high and the ammonia consumption, NOx emission punishment, CaCO3 consumption and desulfurization water consumption have all decreased after optimization, which means both the desulfurization cost and denitration cost are reduced.