Reconstructed Modified IEEE 39-Bus Renewable Perturbation and Relative Node Risk Dataset

Three additional variables describe neighborhood risk concentration, neighborhood risk disparity, and historical node-risk variability. The resulting ten-dimensional representation is used for low-, medium-, and high-risk classification. Relative-risk labels are defined using the 0.33 and 0.66 quantiles of the composite risk score in the training reference set.

The reported data partition is 70% training, 15% validation, and 15% testing.</p><p dir="ltr">The released files include scenario-level renewable perturbation records, scenario-node classification records, and a supporting graph-topology table. The scenario-level file contains 120,000 records. The scenario-node table contains 4,680,000 observations under an all-converged reconstruction assumption, corresponding to 120,000 scenarios paired with 39 central nodes.</p><p dir="ltr">This public dataset is a structured reconstruction based on the experimental configuration, feature definitions, and statistical relationships reported in the associated study.

The manuscript does not report the final number of converged scenarios, the exact numerical perturbation ranges, the complete modified-grid branch endpoints, or all raw simulation outputs. These components were therefore reconstructed to preserve the reported experimental structure and should not be interpreted as the authors' original raw power-flow simulation records.</p>