MACHINE LEARNING AIDED PROCESS DESIGN AND SYSTEM OPTIMIZATION IN WASTE VALORIZATION FOR NEGATIVE CARBON EMISSIONS

Abstract

Waste-to-energy/resource is a sustainable strategy to save traditional energy and protect the environment. However, it is still challenging to fully understand and optimize these waste conversion processes for increasing product quality. Also, no systematic and consistent life-cycle evaluation method was reported for aiding the decision-making of waste valorization. In this thesis, various ML methods were employed to model, interpret, and design both waste-to-biochar and waste-to-energy technologies, including anaerobic digestion, hydrothermal carbonization, hydrothermal liquefication, hydrothermal gasification, slow pyrolysis, and dry gasification. A new ML-based conversion system design and optimization framework was developed to consistently evaluate the energy and emission profiles of waste valorization systems from a holistic life-cycle aspect. The developed ML models have great significance to aid product characterization, process design, catalyst screening, system optimization, and LCA evaluation of waste valorization. Especially, the ML-based process design is beneficial to accelerate the experimental procedure to save labor, time, and cost.