Data-Based Predictive Control for Wastewater Treatment Process

Wastewater treatment process (WWTP) has long been a challenging industrial issue due to its built-in uncertainties and discontinuous measurement of system states. To solve this problem, in this paper, a data-based predictive control (DPC) strategy, based on the available sensing measurements, is proposed to control the dissolved oxygen (DO) concentration in WWTP. First, a self-organizing fuzzy neural network, which can adjust both the structure and parameters simultaneously, is developed to identify the real-time states of WWTP. Second, an improved nonlinear predictive control method is designed to reduce the online computation complexity by transforming the constrained conditions into an unconstrained nonlinear programming problem. Then, an adaptive second-order Levenberg-Marquardt algorithm is developed to derive the control law of DPC. Third, the theoretical analysis on the stability is also given to confirm the prerequisite of any successful application of DPC. Finally, the proposed DPC strategy is applied to the Benchmark Simulation Model No. 1. Experimental results demonstrate that the control performance of the proposed DPC is better than some existing methods.