Machine learning assisted optimization of polyoxometalate catalyzed lignin oxidation and depolymerization through reverse design

doi:10.1016/j.resconrec.2025.108337

Machine learning assisted optimization of polyoxometalate catalyzed lignin oxidation and depolymerization through reverse design

2025-06-01

0

PRE

AI

DOI：10.1016/j.resconrec.2025.108337

Original

Original request for help

Save

Abstract

En 中文

Biomass, as a renewable carbon resource, holds significant potential in addressing the growing energy crisis. Lignin, a key biomass component, can be converted into valuable compounds via polyoxometalate catalysis. Traditional methods require trial-and-error in catalyst selection and reaction optimization, leading to high consumption of chemical reagents and significant waste of resources, energy, and time. To address these issues, we integrated machine learning to develop a predictive modeling and reverse design system for lignin depolymerization, optimizing catalyst selection and reaction conditions. Among the models evaluated, the Artificial Neural Network (ANN) outperformed others, achieving R2 values of 0.91, 0.90, and 0.90 for lignin conversion and phenol/acetophenone yield prediction, respectively. Incorporating virtual sample generation (VSG) and prior knowledge significantly improved prediction accuracy, increasing the average R2 of the models from 0.66 to 0.98. This study provides a more efficient and cost-effective approach to lignin depolymerization, improving resource utilization and contributing to sustainable development.

Keywords:

Machine learning

Lignin depolymerization

Virtual sample generation

Prior knowledge

Catalyst

Machine learning assisted optimization of polyoxometalate catalyzed lignin oxidation and depolymerization through reverse design

Abstract

Journal

Organization