(371a) Machine Learning on Prediction Hydrogen Production from Biomass and Plastic Waste Gasification

Hydrogen currently plays a crucial role as an industrial feedstock and has potential to support decarbonization for industry and to help balance a net-zero energy system. Gasification can be used to produce hydrogen from hydrogen-containing compounds such as fossil fuels, renewable sources (like biomass), and wastes (plastic waste). It is a very complex thermochemical process that involves a series of physical transformations and chemical reactions within gasifier. Gasification performance varies with many factors of feedstock types and properties, gasifier types and their characteristics, operating conditions (such as temperature, gasifying agent, catalyst), etc. In this study, machine learning (ML) was applied to gasification to help understand and predict H₂ production of biomass and plastic gasification. A database of over 800 entries and 40 variables (numerical and categorical) was compiled from literatures on biomass and plastic gasification. The data was pretreated using cleaning and normalization methods. Statistical analysis was performed to determine correlation between variables. The Pycaret library was used and a highly generalizable extra trees linear regression model, with high prediction accuracy (R2 = 0.96 and root mean square error = 4.47) for both training and test data sets was developed to predict H₂ production of plastic and biomass gasification. This ML approach has the potential to aid in optimizing gasification systems for sustainable practices and clean H₂ production.