Machine Learning Prediction of Syngas Composition of Hydrothermal Gasification from Wet Organic Wastes
Jie Li, Lanjia Pan, Manu Suvarna, Yen Wah Tong, Xiaonan Wang
Abstract
Hydrothermal gasification is an effective and economic technology for production of combustible gases and valuable chemicals from wet wastes. In the present work, machine learning (ML), a data-driven approach, is employed to predict the composition of syngas in terms of H2, CH4, CO2, and CO). A gradient boosting regression (GBR) model with optimal hyperparameters was developed for the prediction of syngas composition with a test R 2 of 0.92, 0.90, 0.95, and 0.92 for H2, CH4, CO2, and CO prediction, respectively. This ML framework provides useful model inference, to identify the correlation and causal analytics between the inputs (feedstock compositions and operational conditions of HTG) and outputs (syngas compositions) essential for our future work, and it lays a concrete foundation to devise ML-based process optimization or inverse design for experiments.