State Estimation by Artificial Neural Networks in a Continuous Bioreactor

A based neural networks state observer to estimate biomass, substrate and methane in a continuous anaerobic reactor is introduced in this paper. The observer is designed from a recurrent high order neural network with a hyperbolic tangent as activation function and an extended Kalman filter as learning algorithm. The observer structure is validated via simulations and using experimental data obtained from an anaerobic continuous stirred tank at lab scale. This prototype is used to treat real slaughterhouse wastewater and it is operated in continuous mode. The obtained results show that the proposed observer is able to reproduce adequately the biomethane production and the substrate (related to chemical oxygen demand) in the methanogenesis stage; besides, methanogenic bacteria are also well estimated but some modifications are required in order to reach better results.