Generating Energy and Greenhouse Gas Inventory Data of Activated Carbon Production Using Machine Learning and Kinetic Based Process Simulation

Understanding the environmental implications of activated
carbon
(AC) produced from diverse biomass feedstocks is critical for biomass
screening and process optimization for sustainability. Many studies
have developed Life Cycle Assessment (LCA) for biomass-derived AC.
However, most of them either focused on individual biomass species
with differing process conditions or compared multiple biomass feedstocks
without investigating the impacts of feedstocks and process variations.
Developing LCA for AC from diverse biomass is time-consuming and challenging
due to the lack of process data (e.g., energy and mass balance). This
study addresses these knowledge gaps by developing a modeling framework
that integrates artificial neural network (ANN), a machine learning
approach, and kinetic-based process simulation. The integrated framework
is able to generate Life Cycle Inventory data of AC produced from
73 different types of woody biomass with 250 characterization data
samples. The results show large variations in energy consumption and
GHG emissions across different biomass species (43.4–277 MJ/kg
AC and 3.96–22.0 kg CO2-eq/kg AC). The sensitivity
analysis indicates that biomass composition (e.g., hydrogen and oxygen
content) and process operational conditions (e.g., activation temperature)
have large impacts on energy consumption and GHG emissions associated
with AC production.