Model predictive control for chlorine dosing of drinking water treatment based on support vector machine model

Abstract

Chlorine is the most common disinfectant used in drinking water treatment. To meet the desired disinfection level and restrict the formation of harmful disinfectant by-products, the chlorine dosage should be adjusted in real time to cope with the varying influent water quality and to ensure that the free chlorine residual (FCR) of the clear-water reservoir outlet is within the prescribed limits. This control objective is difficult to achieve by the conventional proportional integral derivative (PID) feedback controls or manual control because of the complicated dynamics of the chlorination process. This study proposes a model predictive control (MPC) scheme for chlorine dosing, in which FCR can be predicted by the support vector machine (SVM) model. Both of the simulation and experimental results show that the proposed MPC scheme has better control performance than the conventional PID feedback control scheme because of the SVM predictions being accurate and the MPC outperforming the PID, and that it can effectively stabilize the quality of treated water.