A new neural observer for an anaerobic bioreactor

Area of Science:

• Biotechnology
• Biochemical Engineering
• Artificial Intelligence

Background:

• Anaerobic digestion processes are crucial for biogas production and wastewater treatment.
• Accurate estimation of key variables like biomass, substrate, and inorganic carbon is essential for optimizing methanogenesis.
• Existing observer methods may face challenges in handling the complex dynamics of these biological systems.

Purpose of the Study:

• To propose a Recurrent High Order Neural Observer (RHONO) for estimating critical methanogenesis variables.
• To apply RHONO for biomass, substrate, and inorganic carbon estimation within a completely stirred tank reactor (CSTR).
• To validate the observer's performance using both simulated and real-world experimental data.

Main Methods:

• Development of a Recurrent High Order Neural Network (RHONN) architecture utilizing a hyperbolic tangent activation function.
• Implementation of an Extended Kalman Filter (EKF) as the learning algorithm for the RHONN.
• Simulation studies to demonstrate the applicability of the RHONO scheme.
• Validation of the observer using data from a laboratory-scale anaerobic digestion process.

Main Results:

• The proposed RHONO successfully estimated biomass, substrate, and inorganic carbon concentrations in the CSTR.
• Both simulation and real-data validation confirmed the observer's accuracy and robustness.
• The RHONN structure with EKF learning proved effective for modeling the complex anaerobic digestion dynamics.

Conclusions:

• The developed RHONO is a viable and effective tool for real-time state estimation in anaerobic digestion processes.
• The observer demonstrates significant potential for implementation in advanced process control strategies.
• This approach offers a promising solution for improving the efficiency and stability of methanogenesis.