RBF neural network predictive control for coagulant dosage

Coagulant control plays an important role in water treatment process. Considering the nonlinear and time-delaying property of coagulation, a model of coagulant dosage based on RBF neural network is developed. The Gaussian function is used for hidden note function, whose centers are adjusted by K-means clustering algorithm. The weights of output layer are obtained based on RLS. The model has been trained and checked by practical data of water plant. The results verify the feasibility of the proposed approach. With optimal algorithm, a closed-loop predictive control system can be established to realize the real time control of optimal coagulant dose rate.