Neural network predictive optimal control for wastewater treatment

This paper presents a two-level controller based on neural network for wastewater treatment process. Predictive optimal control scheme using neural network is applied in the upper level control unit. The optimization unit is designed to generate trajectories of oxygen concentration and heterotrophic concentration. A feed-forward neural network is employed as prediction model. The gradient descent algorithm method is used to realize the optimization procedure. Local control unit contains two PID controllers which maintain the dissolved oxygen concentration and the heterotrophic concentration set-point trajectories. Simulation results demonstrate the effectiveness of the proposed control strategy.