Neural Observer Based Hybrid Intelligent Scheme for Activated Sludge Wastewater Treatment

Activated sludge wastewater treatment plants have received considerable attention due to their efficiency to eliminate biodegradable pollution and their robustness to reject disturbances. Different control strategies have been proposed, but most of these techniques need sensors to measure process main variables. This paper presents a discrete-time recurrent high order neural observer (RHONO) to estimate substrate and biomass concentrations in an activated sludge wastewater treatment plant. The RHONO is trained on-line with an extended Kalman filter (EKF)-based algorithm. Then this observer is associated with a hybrid intelligent system based on fuzzy logic to control the substrate/biomass concentration ratio using the external recycle flow rate and the injected oxygen as control actions. The intelligent system and neural observer performance is illustrated via simulations.