Data and code supporting fast screening of hydro–wind–solar–storage capacity portfolios under network-constrained dispatch

This dataset supports the study “Fast screening of hydro–wind–solar–storage capacity portfolios under network-constrained dispatch using a physics-aware spatial neural operator.” The archive contains processed 2019 hourly renewable-resource, load, and hydropower inputs; definitions for a controlled synthetic eight-node and eleven-branch power system; 120 hydro–wind–solar–storage capacity portfolios; 53 weekly windows per portfolio; and 6,360 main-system scenarios. It also includes fixed train, validation, test, and portfolio-held-out partitions, frozen model checkpoints, numerical source data supporting the manuscript tables and figures, and SHA-256 checksum manifests. External validation data are provided for a stress-enhanced RTS-GMLC Area-1 Lite system containing 24 buses, 38 branches, 40 capacity portfolios, eight weekly windows, and 320 scenarios. The associated reproducibility code is available at: https://github.com/KaangWoo/hydro-wind-solar-storage-screening The main eight-node system is a synthetic research system and does not reproduce confidential Bonneville Power Administration network data. Third-party source data remain subject to their original terms and acknowledgements.