Direct Adaptive Neural Network Control for Dissolved Oxygen Concentration

Abstract

In this paper, a direct adaptive neural-network control (DANC) method is proposed to control the dissolved oxygen (DO) concentration in the wastewater treatment process (WWTP), which employs the forward neural network to approximate an ideal control law. The main features of the DANC system include the following three aspects. Firstly, it is not necessary to establish the accurate plant model, which is favorable because it is an arduous task for the WWTP. Secondly, the multi-condition characteristic of WWTP is considered, and the neural network controller is designed into a self-organizing style. Thirdly, the criterions of growing and pruning network are designed considering the characteristic of WWTP. The adaptive laws of DANC parameters are obtained through the Lyapunov method. Simulation experiments show that the control accuracy and dynamic performance of DO concentration under DANC are improved.