Exact Multi-Source Evaluation for National-Scale EV Charging Station Placement: How Far Does Surrogate Assistance Take Us?

Selecting facility locations on a national road network requires evaluating, for every candidate configuration, the shortest-path distance from each demand node to its nearest facility. We show that this objective admits a single multi-source shortest-path sweep rather than K independent single-source Dijkstra searches, which reduces the cost of one exact evaluation on the Great Britain network (12.8 million nodes, 5.4 million demand points) from an estimated 158 s to about 1 s and makes a fully exact national-scale baseline computationally tractable. We build this exact evaluator and use it as a benchmark for surrogate-assisted multi-objective optimization. We propose GP-SA-MOEA, a Gaussian process surrogate-assisted multi-objective evolutionary algorithm that replaces most expensive distance evaluations with a learned surrogate while correcting it through uncertainty-guided active learning. On a London benchmark with a computable exactly evaluated reference front, GP-SA-MOEA recovers a diverse front (hypervolume ratio 0.995), whereas a pure surrogate without active learning spuriously dominates the reference front by underestimating distances. At national scale, an exactly evaluated NSGA-II baseline built on this evaluator attains a better Pareto front than GP-SA-MOEA at lower cost (mean IGD 556 vs. 4378, in 1.3 h vs. 5.9 h), and the Euclidean proxy is the closer of the two surrogates in aggregate mean. We therefore ask when surrogate assistance is still worthwhile, and answer with a topology-sensitive benchmark: geometric (Euclidean) assignment fails systematically where water barriers and one-way restrictions make network distance diverge from straight-line distance, selecting the wrong facility for every cross-water demand node on both network versions. These findings position the multi-source exact evaluator as a practical route to large-scale multi-objective facility location and delimit the role of surrogate assistance in this setting.