Onshore energy storage for offshore wind in seismically sensitive coastal areas using geographic information systems and machine learning
Journal of Energy Storage, cilt.178, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 178
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.est.2026.123795
- Dergi Adı: Journal of Energy Storage
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
- Anahtar Kelimeler: Autoencoder-based zoning, Battery energy storage systems, Geographic information system, Machine learning, Spatial suitability analysis
- Hacettepe Üniversitesi Adresli: Evet
Özet
This study proposes a fully data-driven geographic information system–machine learning (GIS–ML) framework for identifying optimal land-based battery energy storage system (BESS) sites to support offshore wind energy integration in İzmir, Türkiye, directly contributing to sustainable development goal (SDG) 7. Unlike conventional GIS-based multi-criteria decision-making approaches that rely on predefined weights or expert judgment, the proposed framework employs an unsupervised deep autoencoder (AE) to learn a one-dimensional latent suitability continuum from 16 geospatial criteria evaluated across 138 coastal zones. To enhance robustness and reduce initialization sensitivity, the model was combined with a homogeneous deep ensemble and 200-run bootstrap resampling. The results demonstrate strong structural validity, with high agreement between AE-derived suitability rankings and those independently obtained using uniform manifold approximation and projection (UMAP) (ρ = 0.861, p < 10−40). Ranking reliability was high, as the best-performing zones appeared in the top suitability tier in approximately 82–92% of bootstrap runs, with Zone 6 entering the top 20 in 91.5% of iterations. Permutation-based importance analysis identified proximity to rivers (0.0954), proximity to protected areas (0.0920), and earthquake depth (0.0917) as the dominant suitability drivers, whereas proximity to shorelines showed no discriminative contribution. The practical novelty of the study lies in integrating GIS with unsupervised and bootstrapped deep learning to generate an objective, reproducible, and non-linear zonal ranking without subjective weighting. Compared with traditional GIS-based multi-criteria approaches, the framework better captures hidden spatial interactions, filters out non-informative variables, and provides a more transparent and risk-aware basis for coastal energy infrastructure planning.