Process Integration and Multi-Objective Optimization for Sustainable Global Energy Systems: A Reproducible Framework Coupling $$\varepsilon $$-Constraint Pareto Generation with Multi-Criteria Compromise Selection

Abstract Process integration and optimization methodologies are essential for designing sustainable energy systems that balance environmental, economic, and technical objectives simultaneously. This study develops a reproducible hybrid multi-objective–multi-attribute decision-making computational framework coupling an $$\varepsilon $$ -constraint multi-objective Linear Programming (LP) model with the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) to evaluate sustainable energy mix configurations across diverse national contexts. Using exclusively publicly accessible open-source databases covering 25 countries over 2020–2023, the $$\varepsilon $$ -constraint method generates a verified Pareto frontier of 20 non-dominated solutions spanning the full CO $$_2$$ –cost trade-off space, overcoming the arbitrary weighting inherent in single-objective or fixed-weight scalarisation approaches. TOPSIS subsequently selects the best compromise from this set: a hydro-nuclear portfolio (69.9% hydro, 29.2% nuclear) achieving 0.027 kg CO $$_2$$ /kWh at 79.1 USD/MWh with 55% capacity factor. Structural differentiation from antecedent hybrid frameworks is demonstrated: the $$\varepsilon $$ -constraint formulation guarantees non-dominated solution recovery including non-convex portions, reliability is enforced as a hard constraint within each linear programming model instance, and the multi-criteria method operates exclusively on the verified non-dominated set. Country-level multi-criteria ranking identifies Norway (0.835), Brazil (0.808), and Sweden (0.756) as top performers; sensitivity analysis under ±5% weight perturbation confirms ranking stability. Statistical validation confirms significant inverse correlation between renewable energy share and per-capita emissions (r = $$\varvec{-0.363}$$ , p < 0.05). All data sources and analysis code are fully open, establishing a reproducible and independently verifiable decision-support precedent adaptable to diverse national contexts.