Neural-network-based Decoupling Control for the Post-chlorination Process of Drinking-water Treatment

To maintain the desirable residual chlorine in a clearwater reservoir, optimizing the chlorine dosage, especially in multiple-tunnel multiple-clearwater-reservoir structures, is a challenging task in the post-chlorination process of drinking-water treatment, which involves strong couplings among tunnels, time-varying dynamics, and a long time delay. Focusing on the multi-input multi-output control problem of the post-chlorination process, an intelligent decoupling control method based on inverted decoupling was devised, where the coupling effects are handled by decoupling compensators based on a neural network, and proportional-integral-derivative controllers were designed for the decoupled post-chlorination process. A simulation study revealed that a decoupler based on the Elman neural network can be used to implement the decoupling function. The proposed method is robust when there are disturbances and large model deviations.