Centralized indirect control of an anaerobic digestion bioprocess using recurrent neural identifier

Abstract

The paper proposed to use a Recurrent Neural Network Model (RNNM) and a dynamic Backpropagation learning for centralized identification of an anaerobic digestion bioprocess, carried out in a fixed bed and a recirculation tank of a wastewater treatment system. The anaerobic digestion bioprocess represented a distributed parameter system, described by partial differential equations. The analytical model is simplified to a lumped ordinary system using the orthogonal collocation method, applied in three collocation points, generating data for the neural identification. The obtained neural state and parameter estimations are used to design an indirect sliding mode control of the plant. The graphical simulation results of the digestion wastewater treatment indirect control exhibited a good convergence and precise reference tracking. © Springer-Verlag Berlin Heidelberg 2008.