The role of artificial intelligence in biomass pyrolysis: Advancing predictive modelling and mechanistic understanding through machine learning − A comprehensive review

Biomass pyrolysis is a promising thermochemical pathway for producing renewable fuels and chemicals, yet its optimization remains challenging due to the complex interactions among feedstock properties, operating conditions, and reactor configurations. Artificial intelligence (AI), particularly machine learning (ML), is emerging as a powerful tool to improve predictive modelling and mechanistic understanding of pyrolysis. This review critically examines four pyrolysis regimes (slow, intermediate, fast, and flash) and advanced strategies such as co-pyrolysis, catalytic enhancement, and microwave heating, before discussing how ML algorithms are reshaping data-driven modelling. Supervised models, including random forests (RF), extreme gradient boosting (XGBoost), and artificial neural networks (ANNs), have achieved high accuracy in predicting product yields, while unsupervised and reinforcement learning approaches reveal hidden trends and enable adaptive process control. Furthermore, explainable AI frameworks (XAI), such as Shapley Additive Explanations (SHAP) and partial dependence plots (PDP), provide mechanistic insights that bridge data with reaction pathways. By combining AI with reactor optimization and mechanistic explanation, this field offers a pathway to efficient and scalable pyrolysis systems. Finally, we outline research gaps and future directions, emphasising the need for interpretable and generalisable models that can guide large-scale deployment.