Dynamic Multiobjective Optimization for Wastewater Treatment Using Bayesian Optimization and Model Predictive Control with Observer

Vollständiger Abstract

Worum geht es in dieser Arbeit?

In wastewater treatment processes, there exists a typical conflict between the objectives of energy consumption (EC) and effluent quality (EQ). Achieving energy savings while ensuring effluent compliance remains a critical challenge. This article proposes a dynamic optimal control method to reconcile this trade-off. The approach integrates intelligent modeling, multiobjective optimization, and advanced control techniques. First, a hybrid optimization strategy combining Bayesian optimization (BO) and the multiobjective evolutionary algorithm based on decomposition (MOEA/D) is developed to efficiently search the Pareto front on the Gaussian process regression model. This strategy dynamically generates optimal setpoints for control variables while balancing exploration and evaluation efficiency. Second, a nonlinear model predictive controller (NMPC) enhanced by an extended state observer (ESO) is designed to precisely track these optimal setpoints. The ESO estimates and compensates for internal and external disturbances in real time, thereby improving control robustness. Last, a dynamic multiobjective optimal control strategy is presented, where BO-MOEA/D is utilized to determine the controller setpoints, and NMPC-ESO is employed to execute tracking control. The proposed method is validated on the Benchmark Simulation Model No. 1. Simulation results demonstrate that, compared with conventional strategies, the proposed approach can more effectively identify and track optimal operating points. It significantly reduces overall EC while consistently meeting EQ standards, thereby verifying its effectiveness and superiority for the optimization of complex, nonlinear processes.