The conventional formula estimating method for prediction of SO_2 emission from the desulfurization tower outlet is difficult to be popularized

and it is time-consuming and money-consuming by on-site measurement method. To solve this problem

a prediction method for SO_2 emission from desulfurization tower outlet based on RBF neural network was proposed in this paper. To predict the on-line monitoring data of the dual-alkali desulfurization tower. The results show that

the maximum and average deviation of the RBF neural network are both 0

indicating it is superior to the gradient varying BP neural network and the LM-BP neural network. On this basis

the RBF neural network was used as the prediction model. The input parameters are: the average flue gas temperature is 101 ℃

the average pH value of the desulfurization tower liquid is 8.40

the average liquid-gas ratio is 1.130 0 L/m3. Through continuously changing one of the input parameters

the effects of other input parameters on conversion value of SO_2 mass concentration at the desulfurization tower outlet were investigated. The results show that

the conversion value of SO_2 mass concentration increases with the flue gas temperature

it begins to fluctuate when the flue gas temperature increases to higher than 110 ℃. The conversion value of SO_2 mass concentration decreases with the increasing liquid-gas ratio. When the flue gas temperature is 101 ℃ and the liquid-gas ratio is 1.130 0 L/m3

the optimum pH value of the desulfurization tower liquid is 8.60. Therefore

the above SO_2 emission prediction method for desulfurization tower based on RBF neural network can not only effectively predict the SO_2 emission from the double alkali desulfurization tower outlet

but also can optimize the SO_2 emission reduction.