Design of an Industrial Supervisor for Monitoring and Control

Abstract

In the context of energy transition, anaerobic digestion offers a sustainable way to produce biogas from organic waste. Yet, its biological complexity hinders real-time control, especially due to unmeasurable variables like biomass. This work proposes an industrial supervisor for monitoring and controlling an anaerobic digestion bioreactor, using the AM2 model. Three observers are developed and compared: asymptotic, interval-based, and one using machine learning (Random Forest). Results show each approach offers unique advantages in accuracy, robustness, and uncertainty handling.