Ammonia Spraying Optimization of an SCR Denitrification System Based on RBF Neural Network

To optimize the control on ammonia spraying of the selective catalytic reduction (SCR) denitrification device in a 350 MW power boiler in Guangzhou, a relationship model was established between the input and output variables based on radial basis function (RBF) neural network by taking the boiler load, flue gas flow, SCR inlet flue gas temperature, SCR inlet NO_x concentration and the spraying ammonia flow as the input variables, and the SCR denitrification efficiency as the output variable, so as to realize the prediction of SCR denitrification efficiency and outlet NO_x concentration. Under the premise of satisfying the requirements of NO_x emission and aiming at minimizing the operating cost of the SCR system, Matlab was used to perform a simulation experiment on the model to seek a critical point among the ammonia consumption cost, power consumption cost and the NO_x emission fee, thus obtaining an optimal flow of spraying ammonia. Results show that the calculated mass flow of ammonia spraying is either higher or lower than the measurements, but the operating cost of the SCR system always keeps decreasing under the premise of meeting the NO_x emission standard.