Neural Observer for an Abattoir Wastewater Treatment Process

The paper deals with the development of a Recurrent High Order Neural Observer for an abattoir wastewater treatment by anaerobic digestion. The neural network uses the hyperbolic tangent as activation function and the learning algorithm is based on an extended Kalman filter. This observer is designed to estimate the variables of the methanogenesis stage (biomass and substrate) in a completely stirred tank reactor. A prototype treating real abattoir wastewater is implemented in order to obtain an experimental model, which is used to validate the proposed observer.