An Interaction-Aware NI-EA Framework for EV Charging-Station Siting: Source-Conditioned Robust Candidate Sets and Bounded Spatial Evidence in Dubai

Public-data electric-vehicle charging-station siting needs a screening workflow that can use spatial proxies while keeping demand, grid-capacity, and implementation claims separate from the score. This study develops an interaction-aware Nonlinear Interaction–Einstein Aggregation (NI-EA) framework for Dubai and extends it with source-conditioned robust candidate-set diagnostics. From 7410 admitted candidate/amenity records, 5097 inside-boundary candidates are scored using a candidate-derived activity-density proxy, a charger-coverage-gap proxy, and a grid-access proxy. The analysis compares NI-EA with WSM, the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), and Einstein aggregation; reconstructs a 63-scenario interaction/curvature/blending rank matrix; evaluates weighting, road-network, and official-DEWA source sensitivities; and reports necessary and possible top-K candidate sets, family-balanced finite-scenario acceptability, rank-displacement summaries, and bounded spatial-evidence context from official community, transport, parking, DEWA, and OpenStreetMap-derived sources. The baseline leader is S1421/Boonmax, while official-DEWA coordinate-source reconciliation changes the leader to S3473. Across the reconstructed interaction, weighting, road-network, and official-DEWA scenario families, the top-15 necessary core contains 12 candidates, and the top-15 possible envelope contains 18 candidates. Activity-radius and charger-count coverage alternatives are reported separately as proxy-definition sensitivities. TOPSIS has 0/15 top-15 overlap with NI-EA because it favors a different profile with much higher coverage-gap scores but low activity density. The reported output is therefore a source-conditioned planning shortlist and robustness audit, not an observed-demand map, feeder-capacity validation, financial feasibility assessment, or construction recommendation.