Data-Driven Model Predictive Control Strategy for Coagulant Dosing in Water Treatment Plants

Traditional control methods for the coagulation process in water treatment plants are often inefficient and costly. In this paper, a data-driven coagulant dosage prediction model is introduced using an optimization framework based on genetic algorithm (GA) improved fuzzy C-means clustering (FCM) and adaptive neuro-fuzzy inference system (ANFIS), which can be adapted to the dynamic changes of raw water quality. Through model predictive control (MPC), a multi-objective controller is designed for determining the coagulant dosage set point and control law, integrating optimization and control. After MATLAB/Simulink simulation and practical application in water treatment plants, the method was shown to have significant advantages in cost reduction and water quality stabilization, confirming its potential for coagulant dosage control in water treatment setups.