Optimizing Process Conditions for H2 Production from Biomass and Plastic Waste Gasification using Machine Learning and Genetic Algorithm

Plastic waste and biomass gasification is a promising method for producing renewable hydrogen gas, a useful clean fuel that can be used to transition difficult to decarbonize economic sectors away from fossil fuels.Machine learning (ML) models were utilized to optimize process conditions and feedstock compositions by creating a model from a database with approximately 1000 entries and 24 independent variables, along with various gas compositions such as H 2 and CO.Data was preprocessed and manually condensed to clean entries.Feature engineering methods were used to narrow down these variables to the 6 most important features.A reasonably accurate ML model ( 2 =0.88) was built using the PyCaret package in Python.The effect of reducing feature variables on ML accuracy was also investigated.Genetic algorithm was implemented on the ML model to optimize 2 production, finding that 2 concentration of 0.71 could be obtained without a catalyst.