Advancing Two-Phase Anaerobic Digestion: Operation and Scale-Up Using Kinetic and Machine Learning-Based Approaches

Do-Gyun Kim

Anaerobic digestion (AD) is a vital technology for renewable energy generation and sustainable waste management, yet conventional single-phase systems often exhibit low efficiency and instability when treating high-solid substrate or instability when treating high nitrogen-rich, dairy manure. This dissertation advances a two-phase anaerobic digestion (TPAD) that integrates a hyperthermophilic anaerobic acidification reactor (AAR) with a thermophilic upflow anaerobic sludge blanket (UASB) reactor. The overarching objective is to demonstrate and validate this novel system and to develop hybrid kinetic–machine learning (ML) models to optimize performance and scalability across bench- and pilot-scale operations.The research addresses key challenges in AD— slow process kinetics, ammonia inhibition, process instability, and limited predictive modeling—through an integrated experimental and modeling framework. A biogas stripping–gypsum absorption process was developed for in-situ ammonia recovery, achieving effective nitrogen removal while mitigating methanogenic inhibition and producing ammonium sulfate as a recoverable byproduct. A bench-scale two-phase AD system was first designed, built, and operated to evaluate fundamental process performance under controlled conditions. Insights from the bench-scale operation guided the pilot-scale design and operation of the integrated AAR–UASB system, where hyper-thermophilic acidogenesis (70 °C) enhanced hydrolysis and volatile fatty acid production, and thermophilic methanogenesis (55 °C) achieved stable and efficient methane generation. During the operational phase, fault detection techniques using principal component analysis (PCA) were implemented through the ML platform to support process monitoring and identify significant parameters influencing reactor performance. For process prediction, a hybrid modeling framework was developed by integrating modified kinetic models with artificial neural network (ANN) algorithms. The hybrid model achieved high predictive accuracy for biogas production, providing critical insights into operational parameters and scale-up potential. This research demonstrates a comprehensive approach that unites innovative process design, experimental validation, and data-driven modeling to enhance the efficiency, controllability, and scalability of advanced AD systems. The findings contribute to the development of next-generation two-phase AD technologies capable of achieving higher energy recovery, improved process stability, and integrated nutrient management for sustainable waste-to-energy applications.

Advancing Two-Phase Anaerobic Digestion: Operation and Scale-Up Using Kinetic and Machine Learning-Based Approaches

Files and links (1)

Abstract

Metrics

Details