{"id":"5a1f238a-1806-4ead-9c13-959f2b94f5e7","arxiv_id":"2411.13710","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Simulating a 240-bus Midwest feeder shows 4.16 kV lines violate ampacity limits at 2% EV adoption, while 13.8 kV lines tolerate up to 71% adoption.","lead":"A computer simulation of a real Midwest residential grid shows that low-voltage feeders can overload when only 2% of homes charge electric vehicles at the same time. Higher-voltage networks hold up much longer, which gives utilities a clear lever for planning.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline adoption thresholds (2%/21%/71%) are not reproducible from the described methodology: Section II-B only simulates 20%–100% in 20% steps, so Table III's finer thresholds require an undocumented search procedure and depend on a single unseeded random draw.","rationale":"The reader's weakest-assumption choice (simultaneous full-power charging) is a genuine limitation, and I agree it weakens the practical interpretation of the thresholds. However, my read identifies a more immediate, internally checkable problem: the methodology as written cannot produce the paper's headline numbers. Section II-B describes only five penetration rates in 20% increments, but Table III reports thresholds at 2%, 21%, and 71%. Unless an undocumented search was performed, these values are unsupported. Even if a search was performed, the single random allocation—with no stated seed and no Monte Carlo repetition—means the thresholds may be artifacts of one draw. This is distinct from the simultaneity assumption: it affects whether the reported numbers are reproducible at all, independent of how charging behavior is modeled. The qualitative central claim (lower-voltage feeders overload at lower EV adoption) is plausible and supported by the monotone trends in Tables I and II, so I would not reject the paper. But the specific thresholds should be treated as illustrative scenario outputs pending code/data release and a robustness analysis. This matches the reader's CONDITIONAL verdict, so I leave the verdict unchanged, while adding a concrete reason why the conditionality is necessary. The proposed test—re-running with multiple seeds and documenting the sweep—would settle whether Table III holds or whether the thresholds are noise.","tokens_in":7280,"tokens_out":4861,"duration_ms":50088,"concrete_test":"Request the authors' simulation script or a detailed log of all penetration rates actually run. First, confirm whether any rates other than 20/40/60/80/100% were simulated; without that, Table III cannot be generated from the described method. Second, if the finer sweep exists, rerun the full procedure with 100 different random seeds (e.g., 0–99) and record the first adoption rate at which any violation appears for each voltage level. If the 4.16 kV threshold varies by more than ±5 percentage points across seeds, or if the 13.8 kV threshold sometimes exceeds 100%, the single-point thresholds in Table III are not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's most specific quantitative claims are the EV adoption thresholds in Table III: 2% for 4.16 kV, 21% for 6.9 kV, and 71% for 13.8 kV. These numbers are quoted in the abstract and conclusion as the central actionable finding. However, Section II-B states: 'EVs were allocated across buses based on five adoption rates: 20%, 40%, 60%, 80%, and 100%.' No finer sweep, interpolation rule, or binary search is described anywhere in the methodology. It is therefore impossible to derive 2%, 21%, or 71% from the stated procedure alone. If the authors did run additional simulations (e.g., 1% increments), that step is omitted, making the results non-reproducible without code or data. Even with such a sweep, the allocation uses a single random draw from NumPy's Mersenne Twister with an unspecified seed. At 2% penetration on a 1,120-customer system, only about 22 EVs are placed; which buses receive them can shift the first-violation threshold substantially. The reported thresholds are thus a single sample from a random process, not a statistically supported limit. This is load-bearing because the paper's contribution is framed around these precise adoption limits, and the missing documentation prevents an independent check of the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a simulation-based case study of EV charging impacts on a 240-bus, 1120-customer distribution network in the Midwest U.S. The methodology (Section II) assigns EVs probabilistically to households at 20%–100% adoption rates, adds their rated charger power (5, 10, or 15 kW) to the observed hourly peak load, solves power flow in OpenDSS, and computes line loading and ampacity violations across five voltage levels (4.16–34.5 kV). The results show that line loading and violation counts increase with adoption, with 4.16 kV networks exceeding ampacity at a claimed 2% adoption and 13.8 kV networks supporting up to 71% with 10-kW chargers. The paper concludes that higher voltage levels, line reinforcement, and controlled charging are needed to accommodate EV growth.","tokens_in":7556,"tokens_out":4813,"duration_ms":45849,"significance":"The paper addresses a practical concern for distribution utilities: medium-voltage feeder ampacity under residential EV charging. Its strengths are the use of a real utility feeder model, the systematic sweep of voltage levels and charger capacities, and the direct, non-circular computation of violations from power flow results. If the numerical thresholds were reproducible, they would provide a useful screening benchmark. However, the central quantitative claims, especially the adoption thresholds in Table III, are not reproducible from the described methodology, and the modeling assumptions (simultaneous full-power charging, unspecified reactive power update) limit their generality. The qualitative trend is consistent with standard power-flow logic, but the specific adoption limits should be treated with caution until the methodology is completed.","major_comments":[{"comment":"The thresholds 2%, 21%, and 71% in Table III cannot be derived from the stated simulation procedure, which only runs adoption rates of 20%, 40%, 60%, 80%, and 100%. The authors must document the finer search procedure (e.g., step size, criterion for first violation, any binary search) or provide code and data. Without this, the headline result of the paper is unsupported and not independently verifiable.","section":"Section II-B and Table III"},{"comment":"The EV allocation uses a single unseeded random draw from NumPy's Mersenne Twister. At 2% penetration (approximately 22 EVs across 1,120 customers), the specific buses selected materially determine which lines overload. The reported thresholds are therefore one realization of a stochastic process, not a deterministic property of the system. Please provide the seed, or repeat the allocation many times and report the distribution or confidence intervals for the first-violation thresholds.","section":"Section II-B"},{"comment":"The reactive power update is not specified. The statement 'Once P is updated, the next step is to recalculate Q' is not sufficient to reproduce the simulations: the authors do not state the assumed power factor of the chargers, whether the EV load is modeled as constant power, constant current, or constant impedance, or the formula used for Q. Since line current depends on S = sqrt(P^2+Q^2), the loading values in Tables I and II and the thresholds in Table III depend on this undocumented step.","section":"Section II-C"},{"comment":"The assumption that all EVs charge simultaneously at full rated power for one hour is explicitly stated as a modeling choice, but the abstract and conclusion present the resulting thresholds as general facts (e.g., 'violations occur at just 2% EV adoption'). This is a worst-case bound, not a realistic base-case estimate. The authors should either restrict the conclusions to the simultaneous full-power scenario or add a sensitivity analysis with staggered charging profiles to show how the thresholds change under more realistic charging behavior.","section":"Section II-C and Section IV"}],"minor_comments":[{"comment":"In the 34.5 kV block, the Avg % entry at 100% adoption is listed as '5' without a trailing decimal, while all other entries use one decimal place; this is inconsistent formatting.","section":"Table I"},{"comment":"The text states that the hour of peak average P consumption occurs at 13:00 on July 12, 2017, but Figure 2 shows the 24-hour profile for that date; please verify that the peak time in the figure matches the stated 13:00 value or correct the text.","section":"Section II-A and Figure 2"},{"comment":"The cumulative color-coding legend is described in detail but may be hard to parse in the figure; adding explicit legend labels such as '20% only', '40% (including 20%)', etc., would improve readability.","section":"Figure 3"},{"comment":"Equation (2) defines the violation percentage, but the paper also reports violation counts in Table II and Figure 4. Please clarify that a line is counted once for the count metrics regardless of its violation percentage.","section":"Section II-D"},{"comment":"For the 6.9 kV and 13.8 kV blocks, the Count rows have zero entries for some adoption rates, but the Min/Max/Avg cells are left blank; using a dash or '—' in those cells would make the table more consistent.","section":"Table II"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and addresses a timely topic, but the lack of reproducibility of the central numerical thresholds is a serious issue for publication. The qualitative message is sound and the paper could become acceptable if the authors provide the missing methodology details (finer sweep procedure, seed and statistical treatment of random allocation, and explicit reactive power update) and temper the generality of the conclusions. The novelty is incremental relative to existing EV impact studies, but the multi-voltage comparison on a real feeder is useful as a case study."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look if you do distribution planning, but the headline numbers—2%, 21%, 71%—are not supported by the methodology as written. The paper is a straightforward OpenDSS parameter sweep on a real 240-bus Midwest feeder: five voltage levels, five adoption steps (20–100%), three Level-2 charger powers, one worst-case hour. That combination is genuinely useful as a rule of thumb, and the qualitative trend—lower-voltage feeders overload at much lower EV adoption—is physically sensible and consistent with the power-flow logic. The loading tables (Table I) and violation counts (Table II) are internally consistent.\n\nThe main problem is Table III. The text says EVs were allocated at 20–100% in 20% steps, yet Table III reports first-violation thresholds of 2%, 21%, and 71% with no description of any finer sweep or interpolation. On a 1,120-customer system, 2% is about 22 EVs; a single unseeded random allocation makes that number a sample, not a bound. So the central quantitative claims are not reproducible from the paper alone. The authors either omitted the search procedure or ran results they didn't document. Either way, a referee should ask for it.\n\nMinor issues: the reactive-power update after adding EV load is described as 'recalculate Q' but never specified (constant pf? unity?); the power factor for base loads is randomly picked per customer with unspecified seed; and only one hour (system peak) is simulated, so the thresholds are worst-case scenario outputs, not robust predictions. No code or data are provided, which compounds the reproducibility problem.\n\nNone of this kills the paper. The qualitative result is solid, and the loading-level tables are plausible. But the precise adoption thresholds should be treated as illustrative until the method is documented and multiple random allocations are averaged (or bounds given). I'd send it to review—it's a clean, relevant case study with a fixable reproducibility gap. But I wouldn't cite the thresholds in anything I write without first getting the code or a clear description of the search.","headline":"A useful-but-rough rule of thumb on EV adoption thresholds, let down by an undocumented threshold-finding procedure and a single random allocation; worth reviewing after the method is clarified.","tokens_in":8059,"tokens_out":2075,"would_cite":false,"duration_ms":19954,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-12T15:57:55.678285+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}