Integrating machine learning in catalyst design for sustainable hydrogen from plastic waste

The escalating plastic waste crisis has heightened the need for sustainable, scalable valorization strategies. Catalytic conversion of plastic waste into hydrogen offers dual benefits: waste mitigation and clean fuel generation. However, the variability of plastic feedstock and the complexity of reaction conditions pose significant challenges for designing efficient catalysts. Recent advances in artificial intelligence (AI) and machine learning (ML) are increasingly being employed to optimize process conditions for hydrogen production via electrolysis and traditional thermochemical pathways. ML models, such as neural networks and ensemble methods, have demonstrated high accuracy in predicting hydrogen yields and optimizing parameters for the gasification and pyrolysis of plastic waste. ML is also opening new avenues for accelerating catalyst discovery by enabling rapid prediction of catalyst performance, reaction pathways, and surface interactions. Computational tools and data-driven descriptors are being used to interpret complex catalytic systems and guide the design of more effective catalysts. However, their application to plastic-derived intermediates remains limited. Despite progress, significant gaps persist in applying ML to the unique challenges of plastic waste conversion, including catalyst discovery and the handling of heterogeneous feedstocks. Key limitations include the need for larger, high-quality datasets, improved model interpretability and the integration of domain-specific knowledge with advanced simulation techniques. In this review we critically summarized the current landscape of AI-driven catalyst design focusing on hydrogen production from plastic waste. It identified methodological and practical limitations and proposed a roadmap for integrating AI, domain-specific data, and catalysis simulations to unlock new catalysts for sustainable hydrogen production.