A machine learning based regression methods to predicting syngas composition for plasma gasification system

Abstract

Plasma gasification is considered a promising technology that converts waste into energy through an environmentally friendly process. This research focuses on predicting the outputs of this process, utilizing ML regression techniques. Data from previous studies involving different solid fuels were collected, and four regression techniques Random Forest Regression, Gaussian Process Regression, Decision Tree Regression, and Support Vector Regression were employed to predict the levels of CO2, CO, N-2, O-2, H-2, and CH4 in the plasma gasification process. The experimental dataset was gathered using a microwave gasifier with varying air flow rates (50-100 sL/min) and plasma power (3-6 kW). GPR a coefficient of determination (R-2) values exceeding 0.983 for all outputs and low NRMSE and MSLE values (<0.082 and <0.00123, respectively). RFR also performed well, with R-2 values >0.98 for all outputs except CH4 and CO2. SVR exhibited the least favorable performance, while DTR showed intermediate results. Plasma gasification Machine Learning modeling has proven to enable the prediction of the high-complexity chemical reaction chain in gasification. These models hold promise for applications in simulation environments and can be integrated into microcontroller-based systems for practical use for optimization and control. By this means, the cost of plasma gasification processes would be reduced, and so the plasma gasification system would be more flexible.