Scholarlink

The global demand for sustainable energy solutions has accentuated the potential of biomass gasification for converting waste materials into valuable syngas. The composition of hydrogen in the syngas can be enhanced by efficiently optimizing the gasification process over diverse feedstocks. However, optimization using experimental methods for different feedstocks is highly laborious and time-consuming. This study, therefore, focuses on leveraging machine learning (ML) techniques to optimize the process quickly and efficiently. The data for developing the ML model is acquired using a well-validated Aspen Plus model. The syngas composition data for eight different feedstocks are collected, and eight Artificial Neural Networks (ANN) for each of the eight feedstocks are developed using the collected data to predict H2, CO, CO2, and CH4 concentrations in the syngas. The results indicate an exceptional performance of the developed ANN models, which are further used to optimize the gasification parameters for all eight feedstocks. The main objectives of the optimization formulation are maximization of H2 concentration in the syngas while minimizing the CO, CH4, and CO2 concentrations. Finally, this study exploits the Non-Dominated Sorting Genetic Algorithm (NSGA II) to solve this multi-objective optimization formulation. The Pareto fronts obtained through the NSGA II algorithm indicate diverse and converged solutions.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging Machine Learning for Optimization of Gasification Process Across Diverse Feedstocks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Aswitha Tadepalli,
  • Manchala Rama Sai Anusha,
  • Kishalay Mitra

摘要

The global demand for sustainable energy solutions has accentuated the potential of biomass gasification for converting waste materials into valuable syngas. The composition of hydrogen in the syngas can be enhanced by efficiently optimizing the gasification process over diverse feedstocks. However, optimization using experimental methods for different feedstocks is highly laborious and time-consuming. This study, therefore, focuses on leveraging machine learning (ML) techniques to optimize the process quickly and efficiently. The data for developing the ML model is acquired using a well-validated Aspen Plus model. The syngas composition data for eight different feedstocks are collected, and eight Artificial Neural Networks (ANN) for each of the eight feedstocks are developed using the collected data to predict H2, CO, CO2, and CH4 concentrations in the syngas. The results indicate an exceptional performance of the developed ANN models, which are further used to optimize the gasification parameters for all eight feedstocks. The main objectives of the optimization formulation are maximization of H2 concentration in the syngas while minimizing the CO, CH4, and CO2 concentrations. Finally, this study exploits the Non-Dominated Sorting Genetic Algorithm (NSGA II) to solve this multi-objective optimization formulation. The Pareto fronts obtained through the NSGA II algorithm indicate diverse and converged solutions.