{"id":"4c3d0b6f-27d7-4721-804b-a1565c19f829","arxiv_id":"2606.12791","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A public-data, native-resolution synthetic model of the Korean transmission grid (2,217 buses) traces its layout from OpenStreetMap, calibrates circuit counts to national statistics, and ships a frozen 85 GW power-flow case that converges in one deterministic pass.","lead":"This paper presents a 2,217-bus synthetic model of the Korean power transmission grid, built entirely from public data such as OpenStreetMap power lines and government statistics, and released as a frozen, reproducible power-flow case. A generalist should care because it gives researchers a free, high-resolution test bed for a structurally unusual grid — islanded, with long-distance power transfers into the Seoul area — for planning and decarbonization studies.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"154 kV topology fidelity rests on unvalidated OSM coverage; aggregate national calibration cannot rule out regionally biased gap-filling.","rationale":"After reviewing the full manuscript, I find the reader's identified weakest assumption to be the most load-bearing: the topology's fidelity is inherited from OSM crowd-sourced geometry, and the only external validation is an aggregate national length comparison. This is especially serious for a 'geographically grounded' model because the power-flow results (losses, overloads, voltage profile) are direct consequences of the specific graph, and an aggregate metric cannot detect regionally concentrated coverage gaps. The paper is commendably explicit about limitations (Sec IX), but it does not provide a per-region test against EPSIS, which would be straightforward public-data validation. The abstract/body discrepancy (91 GW vs 85 GW, 2.3% vs 2.6% losses, no overloads vs 15 overloads) is an additional editorial error that undermines trust but does not change the scientific assessment of the underlying model. The frozen operating point is a design choice, not a validity claim; the convergence is expected and not a circular validation because the paper does not use convergence as the primary evidence of realism. The proposed regional circuit-km test would settle whether OSM coverage bias systematically distorts the 154 kV network; until then, the model should be treated as conditionally credible.","tokens_in":14717,"tokens_out":6299,"duration_ms":70186,"concrete_test":"Use EPSIS regional (province-level) transmission circuit-length statistics by voltage class (the paper already uses EPSIS regional figures for other purposes) and compute the model's per-region circuit-km for 154 kV and 345 kV. If any province's ratio deviates from 100% by more than, say, 15% (e.g., Seoul/Busan urban regions under-mapped in OSM), the national agreement is confirmed to be a coincidence of compensating errors, and the gap-filled topology is not geographically faithful. This directly tests whether OSM's coverage bias is spatially uniform, the key unverified assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that the 345/154 kV graph is a faithful reconstruction of the real Korean grid—rests entirely on the fidelity of OpenStreetMap coverage and on the feeder-tag circuit-count estimator. The paper's own sensitivity analysis (Sec IV-D-b) shows the alternative maximum-along-path estimator inflates 345 kV circuit length to 114% vs 106% for the adopted rule, and Sec IX-3 admits that 'individual circuits are approximate' and that the 19/130 gap-filled substation corridors are estimated. This matters because the only external validation is the aggregate EPSIS circuit-km comparison (Fig 5a), a national total. OSM's 154 kV substation coverage is only 79%; if the missing 21% is spatially correlated (e.g., urban underground stations), the gap-fill MST will systematically replace dense urban meshes with long radial links, distorting flows, losses, and overload locations. Because the aggregate national circuit-km can match even with such regional bias, the current 94/100/109% agreement does not establish that the topology is correct at the corridor level. The power-flow results (2.6% losses, 15 overloads, voltage profile) are downstream of this topology and inherit the same uncertainty.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a public-data synthetic model of the Korean transmission system, constructed by reassembling OpenStreetMap/OpenInfraMap power-layer geometry into substation-to-substation corridors, gap-filling unmapped substations with a geographic MST, and calibrating aggregate circuit lengths to EPSIS statistics. The model includes generators, loads, transformers, shunts, FACTS, and HVDC links, serialized in a PSS/E-compatible CSV schema. A solved 85 GW (or 91 GW, depending on the version) high-demand operating point is frozen into the data so that a single Newton–Raphson pass reproduces it. The paper reports aggregate circuit-length agreement (94/100/109% at 765/345/154 kV), power-flow convergence, a qualitative cross-check against KPG-193, and an explicit list of limitations.","tokens_in":14981,"tokens_out":10216,"duration_ms":117635,"significance":"If the inconsistencies noted below are resolved, this would be a valuable contribution: a reproducible, CEII-free, geographically grounded national-scale Korean test system with published data and tooling, addressing a real gap in the public test-system literature. The paper's strengths include its transparent public-data-only construction, the documented sensitivity analysis of the feeder-based circuit-count estimator, the honest statement of limitations in Sec. IX, and the release of the dataset, maps, and scripts. The aggregate circuit-length anchor and the KPG-193 structural comparison are genuine external checks, even if not corridor-level validation.","major_comments":[{"comment":"The manuscript contains two irreconcilable descriptions of the released model. The abstract states 2265 buses, 614 generation/renewable sources (151 GW), 3185 AC circuits, 3413 transformers, and a 91 GW evening peak; the full text and Table I state 2217 buses, 512 sources (144 GW), 3708 AC line circuits, 3324 transformers, and an 85 GW summer-afternoon snapshot. These are not minor typographical differences: the contribution is a dataset, and a reader cannot know which file set corresponds to the paper. Please make the abstract, body, tables, and data release internally consistent.","section":"Abstract vs. Sec. I/Table I"},{"comment":"The 154 kV substation gap-fill accounting is inconsistent. The text reports EPSIS=788, OpenInfraMap≈697 (88%), and a deficit of 91 stations via Eq. (3); however Sec. IV-D-c says 130 unreached substations are gap-filled, and Sec. IV-D-d says the final mapped 154 kV count is about 820 (104%). 697+91=788, not 820; 697+130=827, also not 820. Please reconcile whether 'unreached' means unmapped, unconnected, or something else, and provide a single consistent set of counts for the synthetic substations and the final totals.","section":"Sec. IV-D-c and IV-D-d"},{"comment":"The central fidelity claim rests on OSM coverage that is 79% by line-length and 88% by substation count at 154 kV, with 130 substation connections gap-filled by an MST. The only external validation is national aggregate circuit-km (Fig. 5a), which cannot detect regionally correlated bias: if the missing substations are concentrated in urban areas, the MST will replace meshed urban subtransmission with long radial links, distorting flows, losses, and overload locations. Please add a spatial or regional validation (e.g., provincial circuit-km, line-length distributions, or a sensitivity analysis that perturbs the gap-filled corridors) or temper the claim that the layout is a faithful OSM-derived reconstruction rather than a geographically inspired synthetic approximation.","section":"Sec. IV-D and IX-3"}],"minor_comments":[{"comment":"The interactive map URL in Ref. [24] contains 'GIST-2064-bus-test-system', inconsistent with the 2217/2265-bus model nomenclature; update the link or the naming.","section":"Ref. [24]"},{"comment":"Typo: 'V ogel' should be 'Vogel' (sunflower phyllotaxis). Also clarify whether the 1.15 routing/detour factor in Sec. IV-A is a public-data-derived value or a conventional engineering assumption.","section":"Sec. IV-D-d"},{"comment":"The phrase 'coverage is 100%, 96%, and 79% at 765/345/154 kV' is ambiguous: is this line-geometry length coverage, substation coverage, or circuit-count coverage? The later text cites 88% substation coverage at 154 kV; please define both metrics explicitly.","section":"Sec. IV-D"},{"comment":"The single-pass reproduction of the frozen operating point is, as Sec. VI-F acknowledges, a property of baking the solution into the data. Consider stating this more explicitly in the abstract or validation section so readers do not interpret it as an independent validation of the operating point.","section":"Sec. VI-F/VIII"}],"recommendation":"major_revision","confidential_remarks":"The paper describes a useful and potentially citable test system, and the author is commendably transparent about limitations. However, the abstract/body numerical discrepancies and the inconsistent gap-fill counts are serious in a dataset paper. They are fixable within the manuscript's scope; I would support acceptance after a careful revision that reconciles all numbers and strengthens the topology validation or qualifies the fidelity claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe thing to know: this is the first public native-resolution model of the Korean transmission system, and it is assembled with unusual care from public data. It deserves a serious referee, but the manuscript has several fixable problems, and the authors oversell the solved snapshot.\n\nWhat's genuinely new: the multi-source shortest-path reassembly of fragmented OSM line geometry into substation-to-substation circuits, the feeder-based circuit counting, and the order-of-magnitude resolution jump over KPG-193 (2217 vs 193 buses). The 94/100/109% circuit-km calibration against EPSIS is a real external anchor. The paper is also commendably explicit about what is measured and what is estimated (Sec. IX lists shunt set points, tap positions, gap-filled corridors, per-circuit counts). That honesty makes the artifact usable.\n\nThe load-bearing claim—that the 345/154 kV graph is a faithful reconstruction—rests on OSM coverage (79% of 154 kV substations) and on the feeder-tag estimator for circuit multiplicity. The authors' own sensitivity analysis shows the alternative estimator changes 345 kV length from 106% to 114%. Aggregate national circuit-km can match even with regionally biased gap-filling (e.g., urban underground stations missing, replaced by MST radials). So the stress-test concern is legitimate: the topology is plausible, not proven corridor-by-corridor. The paper admits this, but the conclusion states it more strongly than the evidence supports.\n\nThe bigger issue is circularity. The frozen operating point is the output of the authors' own offline settlement loop, baked into the data. So 'a single deterministic pass reproduces the snapshot' is true by construction, and convergence with reactive limits enforced is the fitted target, not independent validation. Sec. VIII-b should be relabeled as consistency, not evidence.\n\nAlso, the abstract contradicts the body on multiple numbers: bus count 2265 vs 2217, generation 151 vs 144 GW, operating point 91 GW evening peak vs 85 GW high demand, losses 2.3% vs 2.6%, 'no overloads' vs 15 circuits >100%. This looks like the abstract was written for a different version. That must be fixed.\n\nNone of this breaks the central deliverable. The dataset, if it runs as described, is a real contribution to reproducible work on an unusual grid. My recommendation: send it to peer review—after the authors reconcile the abstract, soften the convergence-as-validation claim, and ideally archive a versioned DOI with the Overpass queries and one generated figure. The calibration to EPSIS is a nice anchor, but it is aggregate; the paper should not pretend to corridor-level fidelity.\n\nI'd cite it as a test system in my own work and probably bring it to a reading group, mainly to see whether the dataset actually solves as claimed.","headline":"First public native-resolution Korean grid model, honestly assembled from public data, but the solved snapshot is a fitted artifact and the abstract contradicts the body; still deserves a serious referee.","tokens_in":15551,"tokens_out":2306,"would_cite":true,"duration_ms":25575,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Using only public map and statistical data, the paper reconstructs a working native-resolution test system of the Korean transmission grid that reproduces the same 85 GW solved operating point on every run.","keywords":["synthetic power grid","Korean transmission system","public-data modeling","topology reconstruction","power flow test case","frozen operating point","geographically grounded model","CEII-free dataset"],"falsifier":"Re-run the published deterministic power-flow computation on the distributed CSV files; if the reported operating point (voltages, losses, overloaded circuits) is not reproduced exactly, the frozen-operating-point claim fails. Alternatively, compare the model's 345 kV corridors and circuit multiplicities against a reliable public transmission map; any invented or missing trunk corridor would disprove the geographic-grounding claim.","tokens_in":14510,"feed_emoji":"⚡","tokens_out":8183,"duration_ms":87527,"temperature":0.7,"pith_summary":"The paper sets out to prove that a national power grid can be reconstructed at native resolution from strictly public data, using Korea's distinctive islanded transmission system as the test case. It derives the 345 and 154 kV network topology from the geometry of crowd-sourced map data, reassembling fragmented line segments into substation-to-substation circuits and counting circuits from the first segment leaving each substation; unreachable substations are gap-filled with a geographic minimum spanning tree. The aggregate circuit length lands at 94/100/109 percent of published national figures at 765/345/154 kV. The payoff is a 2,217-bus test system with 512 generation sources and 3,708 AC circuits that is distributed with a frozen operating point, so a single deterministic power-flow pass reproduces an 85 GW high-demand snapshot with a 2.6 percent loss level and a transmission voltage profile inside operating limits. A sympathetic reader would care because this supplies the first reproducible, confidential-data-free Korean case detailed enough for AC power-flow, contingency, and planning research.","feed_headline":"Korean grid modeled from public data reproduces an 85 GW snapshot","feed_subtitle":"A reproducible, confidential-data-free case for power-flow, contingency, and decarbonization studies.","key_machinery":"The two mechanisms that carry the argument are (1) the multi-source shortest-path reassembly that converts fragmented map line geometry into a usable substation-to-substation graph, together with the feeder-based circuit-count estimate (using the first segment's tag rather than a path maximum) and the geographic minimum-spanning-tree gap-fill for uncovered substations; and (2) the frozen operating point, where an offline coupled settlement of on-load tap changers, switched shunts, and remote voltage-control gains is baked into the data, collapsing all control loops into a single deterministic power-flow pass. Together they turn crowd-sourced geometry and aggregate statistics into a reproduci","core_discovery":"The central claim is that the real substation-to-substation transmission graph of Korea can be recovered from public geographic data without confidential operator information. The method uses a multi-source shortest-path algorithm to reassemble line fragments into corridors, a feeder-based rule to estimate how many circuits each corridor carries, and a geographic minimum-spanning-tree fallback for the roughly 19 (345 kV) and 130 (154 kV) substations the map data does not reach. The paper then freezes an offline-settled operating point into the dataset—tap ratios, generator set points, and bus voltages—so a single Newton–Raphson computation with reactive-limit enforcement converges identicall","pith_inferences":["A reader might infer that the same reassembly pipeline could be re-run as map data improves, yielding refreshed models automatically—turning a crowd-sourced map into a living grid representation—rather than a one-time snapshot.","Because per-corridor circuit counts are the weakest link, a natural sensitivity study would perturb only the counts and measure how much the 2.6 percent loss figure and the set of overloaded circuits move; large swings would indicate the operating point is partly an artifact of the counting rule.","The paper's 'structurally consistent with an independent coarser model' check could be sharpened into a quantitative cross-validation: compare region-level net imports and transfer flows across the two models, not just qualitative patterns.","The frozen operating point is a convenience but also a limitation: the optional online control loops show how the model would behave if taps and shunts were free to act, so extending the analysis to a time series would require re-freezing at every hour rather than relying on a single point."],"forward_implications":["Reproducible AC power-flow, contingency, and reactive-planning studies on the Korean system become possible without confidential operator data.","The reconstruction methodology provides a template for building geographically grounded synthetic models of other grids from public map data.","Distributing a solved, frozen operating point removes run-to-run nondeterminism, so benchmark comparisons between studies are meaningful.","The explicit 154 kV and 22.9 kV layers make the model a natural transmission-side anchor for combined transmission–distribution studies.","The close match to published circuit-length statistics (94/100/109 percent at 765/345/154 kV) indicates the recovered topology captures the real amount of transmission, even if individual circuits are approximate."],"fun_headline_variants":["Public data reconstructs Korea's transmission grid","OpenStreetMap yields realistic Korean grid model","91 GW Korea grid model built from public maps","Public-data model reproduces Korea's 91 GW snapshot"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The entire topology fidelity rests on the crowd-sourced geographic database being a faithful representation of the real Korean transmission network—especially the 154 kV layer, where only about 79 percent of substations are mapped, and the per-corridor circuit counts inferred from the first line segment's tag.","fun_headline_variants_meta":{"raw":{"variants":["Public data reconstructs Korea's transmission grid","OpenStreetMap yields realistic Korean grid model","91 GW Korea grid model built from public maps","Public-data model reproduces Korea's 91 GW snapshot"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000959,"raw_usage":{"total_tokens":3984,"prompt_tokens":870,"completion_tokens":3114,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":614,"completion_tokens_details":{"reasoning_tokens":3055}},"tokens_in":614,"tokens_out":3114,"duration_ms":21988,"temperature":1.0,"reasoning_tokens":3055,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T11:41:56.656628+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the published deterministic power-flow computation on the distributed CSV files; if the reported operating point (voltages, losses, overloaded circuits) is not reproduced exactly, the frozen-operating-point claim fails. Alternatively, compare the model's 345 kV corridors and circuit multiplicities against a reliable public transmission map; any invented or missing trunk corridor would disprove the geographic-grounding claim.","supporting_citations":[],"review_version":2}