Predicting hydrogen production from co-gasification of biomass and plastics using tree based machine learning algorithms

Abstract

Hydrogen production from co-gasification of biomass and plastics are predicted using Machine Learning Algorithms, e.g., Decision tree and Ensemble methods. Independent variables are particle sizes of biomass and plastics, feedstock ratio and temperatures. The dependent variable is Hydrogen production. Model and prediction performances were evaluated/validated using model parameters. The relative importance scores for independent variables are RSS particle size > HDPE particle size > Temperature > Percent plastics. Size dependence of Hydrogen production indicated a surface-controlled reaction. Temperatures between 500 °C and 900 °C have less impact on H2 production compared to the size. Predictions were carried out using Train-test split, Cross-validation, and GridsearchCV model on the data unseen. Gradient Boosting performed the best.