Whole-Life Carbon–Cost–Circularity Optimization for AI-Era Data Centers: A Robust Multi-Objective Framework Under Carbon Pricing Uncertainty
Buildings · 24.07.2026 · 16 (15) · 2961
DOI: 10.3390/buildings16152961 ↗Kurz gefasst
Fachartikel in der Zeitschrift Buildings, online erschienen am 24.07.2026. Verfasst von Arezou Shafaghat und 2 weiteren Autor:innen. Die Arbeit nennt 57 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The explosive growth of artificial intelligence workloads has triggered unprecedented data center construction, yet environmental assessment remains focused on operational energy efficiency. This paper introduces DC-WLC3O, a framework that simultaneously optimizes three objectives across the full data center lifecycle: whole-life cost, whole-life carbon, and material circularity index. The framework integrates cradle-to-cradle lifecycle assessment with dynamic material flow analysis, employing a four-layer analytical stack: Monte Carlo simulation with Latin Hypercube Sampling and Iman–Conover rank correlation; NSGA-III tri-objective Pareto optimization; Sobol’ variance-based global sensitivity analysis; and Many-Objective Robust Decision Making (MORDM). The framework evaluates 192 design configurations for a 30 MW hyperscale facility under three carbon pricing scenarios. Results reveal a carbon rebound paradox: operational carbon dominates grid-dependent configurations (96.2%), but low-carbon energy produces an 8.5× inversion (Welch’s t = 22.2, p < 10−39). Sobol’ analysis confirms energy strategy dominates carbon outcomes (S1 = 0.87). MORDM identifies on-site renewable energy with a full circular economy strategy as unconditionally robust (R = 0.056, stable across 94.2% of weight permutations). The cost–circularity correlation is negative (rho = −0.488, p < 10−12), overturning the assumption that circular strategies increase lifecycle costs. A post hoc stress test confirms this finding is robust to moderate perturbations in renewable LCOE (+71%), with stability declining to ≈80% under worst-case capital cost assumptions and to ≈61% when generation degradation and battery replacement are jointly considered.
Zitieren
APA
Shafaghat, A., Hu, D., & Keyvanfar, A. (2026). Whole-Life Carbon–Cost–Circularity Optimization for AI-Era Data Centers: A Robust Multi-Objective Framework Under Carbon Pricing Uncertainty. Buildings, 16(15), 2961. https://doi.org/10.3390/buildings16152961
Vancouver
Shafaghat A, Hu D, Keyvanfar A. Whole-Life Carbon–Cost–Circularity Optimization for AI-Era Data Centers: A Robust Multi-Objective Framework Under Carbon Pricing Uncertainty. Buildings. 2026;16(15):2961. doi:10.3390/buildings16152961
BibTeX
@article{Shafaghat2026,
author = {Shafaghat, Arezou and Hu, Da and Keyvanfar, Ali},
title = {Whole-Life Carbon–Cost–Circularity Optimization for AI-Era Data Centers: A Robust Multi-Objective Framework Under Carbon Pricing Uncertainty},
journal = {Buildings},
year = {2026},
volume = {16},
number = {15},
pages = {2961},
doi = {10.3390/buildings16152961}
}
EUVIMED