Artificial Intelligence and Machine Learning in Thermodynamic Gas Hydrate Studies: A Review

Artificial intelligence and machine learning (ML) have emerged as transformative tools for predicting the thermodynamic conditions of gas hydrate systems, offering an efficient and scalable alternative to traditional modeling methods. This review systematically analyzes studies published between 2016 and 2025 using a PRISMA-guided methodology and bibliometric mapping. The analysis reveals a dominant focus on binary water–gas systems (87.5%) and a strong reliance on experimental literature-derived data sets (73.9%). Artificial neural network, support vector regression, and random forest models are the most prevalent algorithms, with R2 and root mean square error as the primary evaluation metrics, though the lack of uncertainty quantification remains a significant limitation. Field-derived data sets are critically under-represented, underscoring the need for standardization, open-access repositories, and industry–academia collaboration. Notably, the review identifies a methodological gap in evaluating model robustness and highlights opportunities for expanding model outputs to include hydrate kinetics and morphology. By integrating bibliometric insights with qualitative analysis, this review not only charts the trajectory of ML applications in gas hydrate research but also provides actionable recommendations for future work, positioning data-driven hydrate prediction at the forefront of energy and environmental innovation.