Multivariable Adaptive Predictive Control for Wastewater Treatment Process Based on ESN

Due to the highly nonlinear and time delay characteristics of wastewater treatment processes,a kind of multivariable adaptive predictive control system based on the echo state network (ESN) model is proposed. First,an intelligent predictive model is established using ESN to predict the outputs of the wastewater treatment process. Second,the ESN identifier is established in order to compensate for error generated by the differences between the actual outputs and the identifier outputs. Finally,experiments are designed based on the BSM1. The proposed multivariable adaptive control strategy is used to control the dissolved oxygen concentration and nitrate concentration. The experimental results show that this control method improves the adaptability and anti-interference ability,and achieves rapid and accurate tracking of dissolved oxygen concentration and nitrate concentration.