Modeling of sustainable methanol production via integrated co-gasification of rice husk and plastic coupled with its prediction and optimization using machine learning and statistical-based models

Abstract

To reduce reliance on fossil fuels and mitigate environmental impact, co-gasification of waste materials presents a promising alternative for methanol production. In modeling gasification process, kinetic-based models are predominant but are often complex and lack inherent optimization capabilities. This study couples a kinetic-based model with predictive models, aiming to provide an optimization-embedded and simplified simulation approach. Using Aspen Plus, an integrated model for methanol production via co-gasification of rice husk and plastic was developed. Model prediction and optimization were performed using response surface methodology (RSM) as a statistical approach and artificial neural network-genetic algorithm (ANN-GA) as a machine learning approach. Key input variables, including gasification temperature (GT), steam-to-feed ratio (STF), methanol production temperature (T) and pressure (P), were optimized for both the co-gasification and methanol sections. The integrated co-gasification-methanol model was successfully developed, achieving a root mean square error (RMSE) of 2.31 when validated with experimental data. Predictions using both ANN-GA and RSM methods yielded a coefficient of determination (R²) > 0.99, with ANN-GA showing superior prediction accuracy. Statistical analysis of variance (ANOVA) from the RSM results also confirmed the model significance. The optimal methanol yield was 0.6 kg/kg feed under GT = 850 °C, STF = 0.96–1.73, T = 234–255 °C, and P = 114–150 bar. While ANN-GA provided superior optimization across most variables, RSM was more effective for optimizing pressure. These findings demonstrate the effectiveness of integrating machine learning and statistical models with kinetic-based simulations for optimizing an integrated gasification-methanol system.