Balancing energy security and sustainability in G20 countries: a hybrid fuzzy Delphi multi-criteria framework

This study examines the tension between energy security and energy sustainability across the nineteen sovereign members of the Group of Twenty (G20). The framework is novel in three respects. First, criterion weights are elicited from a panel of six large language models (LLMs) acting as an exploratory expert surrogate, through a documented, consistency-audited, two-round hesitant fuzzy Delphi procedure whose prompts, model versions, parameters, and raw outputs are fully archived. Second, these subjective weights are fused multiplicatively with an equal-credibility ensemble of three objective weighting methods—Criteria Importance Through Intercriteria Correlation (CRITIC), Logarithmic Percentage Change-driven Objective Weighting (LOPCOW), and Objective Pairwise Adjusted Ratio Analysis (OPARA)—computed from real indicator data for 2010–2022. Third, countries are ranked with five multi-criteria decision-making methods—the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), Multi-Objective Optimization based on Ratio Analysis (MOORA), VIseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR), Weighted Aggregated Sum Product Assessment (WASPAS), and Complex Proportional Assessment (COPRAS)—, whose outcomes agree closely (pairwise Spearman rank correlations of 0.73–0.99) and are aggregated into a single Borda-count consensus, and a formal trade-off index—the difference between a country’s normalized security and sustainability scores—classifies the economies into four policy profiles. Resource-endowed exporters (Australia, Canada, Saudi Arabia) dominate the security dimension, whereas economies combining low carbon intensity, high energy efficiency, and high renewable shares (Brazil, Italy, Germany, the United Kingdom, France) lead the sustainability dimension; India and South Africa are weak in both dimensions, and China, the largest producer of renewable-energy patents, ranks in the lower-middle range because high carbon and energy intensity offset its innovation output. An extensive robustness battery—eight weighting schemes, leave-one-model-out and single-rating perturbation of the elicitation, common-period, rolling-window, pre/post-2020, indicator-substitution, and patent-normalization analyses—leaves the consensus rankings essentially unchanged, and every result is reproducible from the archived inputs with the supplied code. The framework offers policymakers an auditable decision-support tool for navigating the security–sustainability trade-off.