{"id":"2b530be6-719e-422f-b32b-7fa096b6aec1","arxiv_id":"2506.02234","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A reformulation of the SOC voltage cone for optimal power shutoff cuts solve times dramatically, while full linearizations overestimate deliverable load and do not beat DC power flow.","lead":"This paper tests faster ways to plan wildfire-related power shutoffs while keeping as many customers served as possible. It finds that a simple rewrite of the voltage equations cuts solve time from hours to under a minute on medium grids, and that fully linear approximations overpromise how much power can be delivered.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The negative claim about linearizations rests on SOC-redispatch as a proxy for AC deliverability; this proxy is unvalidated and could change cross-method rankings, though the SOC-OPS-P speedup itself looks sound.","rationale":"The reader's weakest assumption is the load-bearing one for the paper's negative result. The positive contribution, the SOC-OPS-P reformulation, appears mathematically sound: (17) is equivalent to the original voltage cone (15) on the integer feasible set and gives a tighter continuous relaxation, because W_Fr<=W_ii, W_To<=z_ij*ub(W_jj), and symmetrically, so W_Fr*W_To is no larger than each right-hand side in (15). The case-57 speedup is also qualitatively robust even though Table I's 13,530 s average includes one time-limited scenario: the other four scenarios average roughly 3,400 s versus 45.75 s for SOC-OPS-P. The concern is specifically about the evaluation of solution quality for the linear relaxations. Table III measures redispatch performance with SOC power flow, not AC, and the paper's own framing is about AC-feasible load delivery. Since SOC is a relaxation, SOC-redispatch performance is at best an optimistic proxy; without an AC redispatch check, absolute statements about overestimation and especially the comparison 'linearized solutions are no better than DC-OPS' are not fully established. This does not invalidate the paper, but it justifies maintaining the reader's conditional verdict rather than upgrading to acceptance. An AC redispatch of the fixed topologies is a direct, inexpensive check that would settle whether the proxy changes the conclusions.","tokens_in":11852,"tokens_out":22837,"duration_ms":213732,"concrete_test":"Reproduce Table III using AC redispatch: for the 57-bus and 89-bus cases, fix the z variables returned by SOC-OPS, SOC-OPS-M, SOC-OPS-S, and DC-OPS, then solve a full AC OPF with continuous load delivery (e.g., via PowerModels AC) and compute the AC-achieved load divided by each formulation's predicted load. If the AC ratios and relative rankings match Table III within a small tolerance, the proxy concern is resolved; if they differ materially, the negative conclusion must be restated as SOC-proxy overestimation rather than AC-feasible overestimation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central negative conclusion—that full linearizations overestimate deliverable load and are no better than DC-OPS—is evaluated in Table III by fixing the binary variables and re-solving with the SOC-Redispatch problem from [6] (Section IV), not by solving the full nonlinear AC power flow. The abstract and introduction motivate the work in terms of AC-feasible load delivery, so the evaluation metric is a proxy for the thing the conclusion claims to measure. The SOC voltage cone is a relaxation of AC, so SOC-Redispatch can predict more deliverable load than AC; moreover, because each row of Table III is a ratio to a different SOC-redispatch denominator, a lower ratio for one formulation and a higher ratio for another does not establish which topology actually delivers more AC load. The overestimation conclusion is directionally safe only in cells where the ratio is below 100% (predicted load exceeds the SOC upper bound, and hence also AC), but cells near 100% and the relative rankings among Secant, McCormick, and DC formulations are not settled without an AC check. The SOC-OPS-P reformulation itself appears exact for binary feasible sets: for z_ij=1 the constraints force W_Fr=W_ii and W_To=W_jj, so (17) matches (15a); for z_ij=0 all voltage variables vanish. Its speedup contribution is therefore not affected by this concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies topology optimization for power shutoff planning under large-scale outages, comparing a SOC-based optimal power shutoff (OPS) formulation with several linearized alternatives. The authors propose an exact reformulation of the voltage cone as a single rotated SOC constraint linking the line-end voltage variables (Eq. 17), and report that this reformulation reduces solve time dramatically on several PGLib cases. They then develop linear relaxations of the thermal limits and of the voltage cone, using tangent and secant cuts, and evaluate the resulting formulations on PGLib networks from 14 to 118 buses. The paper concludes that the SOC-OPS-P reformulation is a major practical speedup, while the fully linearized models overestimate deliverable load and do not provide a better speed-quality trade-off than DC-OPS.","tokens_in":12161,"tokens_out":10207,"duration_ms":97969,"significance":"The SOC-OPS-P reformulation, if correct, is a valuable and easily reusable modeling improvement for topology-switching SOC power flow, with reported speedups of two to three orders of magnitude on the 57-bus and 60-bus cases (Table I). The paper also makes a useful negative claim: replacing the SOC voltage cone by simple linear cuts does not preserve deliverable load quality. These claims are supported by experiments on public PGLib cases with multiple scenarios. However, the definition of the McCormick-based linear model contains a substantive error, and the evaluation of load deliverability is based on SOC redispatch rather than AC power flow, so the negative conclusions require correction and additional validation before the paper can be accepted.","major_comments":[{"comment":"The model SOC-OPS-M is claimed to be fully linear, but Eqs. (23a) and (23b) contain the products W_Fr W_To and W_Fr W_To z_ij, which are bilinear and trilinear, respectively. Moreover, the two constraints are identical as printed, and neither resembles a McCormick envelope for the bilinear term W_Fr W_To. Since this is the defining model for the McCormick relaxation, the numerical results for SOC-OPS-M in Tables I-III rest on an undefined model. Please replace Eqs. (23a)-(23b) with the actual McCormick envelopes using the bounds on W_Fr and W_To (or the intended disjunctive formulation), and then re-run or re-report the affected experiments.","section":"Section III-B, Eq. (23)"},{"comment":"The thermal-limit cut for the reverse direction is written as y_P_ji + y_Q_ji <= T_ij z_ij, but the original thermal constraint (8) has the squared limit T_ij^2 z_ij. The square is missing in (21a). As printed, this constraint is not a valid relaxation of (8): for T_ij > 1 it is tighter than the thermal limit and may cut off feasible flows, while for T_ij < 1 it is too loose. Since the SOC-OPS-T, SOC-OPS-M, and SOC-OPS-S models all include (21a), their feasibility and optimality properties are affected.","section":"Section III-A, Eq. (21a)"},{"comment":"The paper evaluates solution quality by fixing binary decisions and re-solving with SOC-Redispatch from [6], not by solving the AC power flow. The abstract and introduction motivate the work through AC-feasible load delivery and describe the negative result as an overestimation of deliverable load. Because the SOC voltage cone is a relaxation of AC, SOC-Redispatch can report higher deliverable load than AC; consequently, the relative rankings among Secant, McCormick, and DC formulations in Table III are not established for AC deliverability. This is especially important for cells near 100% and for the comparison between Secant and DC. Please either run an AC-feasible redispatch for the fixed topologies or clearly qualify the negative conclusions as statements about SOC-feasible load delivery rather than AC-feasible load delivery.","section":"Section IV and Table III"},{"comment":"The paper claims that when z_ij is relaxed to be continuous, the right-hand side W_Fr W_To in (17) is never larger than any of the right-hand-side terms in (15). This claim is false in the continuous relaxation. For example, take z_ij = 0.5, W_ii = W_jj = 1, and V_i^2 = V_j^2 = 0.9, Vbar_i^2 = Vbar_j^2 = 2. Then W_Fr = W_To = 1 is feasible for (11)-(12), giving W_Fr W_To = 1, while W_ii W_jj z_ij = 0.5. Thus (17) is not tighter than (15) in the continuous relaxation. The reformulation is exact on binary feasible sets, which is sufficient for the main speedup claim, but the stronger continuous-relaxation statement should be removed or replaced with a precise proof.","section":"Section II-B, Eq. (17)"}],"minor_comments":[{"comment":"Equations (15b) and (15c) are identical as printed, with the same right-hand side W_ii W_jj z_ij. This appears to be a typo; please clarify the intended three constraints or remove the duplicate.","section":"Eq. (15)"},{"comment":"The text in Section V-C describes a case study on the 73-bus system, but Figure 2 is titled 'SOC-OPS-M on the 89 bus system'. Please align the text and the figure.","section":"Section V-C and Figure 2"},{"comment":"The discussion states that SOC-OPS-T reduces the number of SOC constraints from 3|B| to |B|, but the relevant count is per line, so it should read 3|L| to |L| (or the sentence should be reworded accordingly).","section":"Section VI"},{"comment":"The variable y_WSum is introduced as an upper bound for the squared term [1/2(W_To - W_Fr)]^2, but the cut in Eq. (24) is written directly in terms of W_To - W_Fr. Please clarify the scaling of the linearization points l and the relation between l and the argument of the square, to avoid dimensional ambiguity.","section":"Section III-C, Eq. (24)"},{"comment":"The introduction says the goal is to find good binary decisions and then solve an AC power flow continuous problem, but the evaluation in Section IV uses SOC-Redispatch. Please align the terminology so that the evaluation metric is consistently described as SOC-based, or change the implementation.","section":"Introduction and Section IV"}],"recommendation":"major_revision","confidential_remarks":"The SOC-OPS-P speedup is the strongest and most credible contribution, and it appears to be well supported by the experiments. The negative results on linear relaxations are potentially interesting, but they are currently undermined by the invalid definition of the McCormick model and by the use of SOC redispatch as a proxy for AC deliverability. I recommend that the editor require the authors to correct the model definitions, add an AC-based redispatch check or substantially qualify the deliverability claims, and resubmit. I do not see the issues as irreparable within the scope of the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the exact reformulation in Section II-B is real, the speedup is real, and the negative result about full linearizations is probably right in the large, but not proven as carefully as the abstract implies. The paper deserves serious review, but needs fixes.\n\nThe new thing is the voltage-cone rewrite (W_R)^2+(W_I)^2 <= W_Fr*W_To, which is exact when the binary switch is on and forces everything to zero when off. That is a clean observation, and the runtime gain on the 57-bus case (13,530s to 45.75s) is striking. The paper does not oversell it: it notes the reformulation does not help beyond 60 buses. Also good: the negative result about linear relaxations is framed honestly. The authors tried two linearization schemes and report that they overestimate deliverable load and do not beat DC-OPS. That is useful for people deciding where to spend effort.\n\nWhere it gets soft: the evaluation metric is SOC redispatch, not AC. The abstract and intro lean on 'AC-feasible', but the tables are ratios against SOC-redispatch denominators. For cells well below 100%, the overestimation conclusion is directionally safe. But the relative rankings among Secant, McCormick, and DC on cases where ratios are close (case39, case60) could shift if checked against AC. That is a real gap, and the paper should either add AC validation or soften the language.\n\nThere are also two text errors. Eq. (21a) writes the thermal limit as T_ij z_ij instead of T_ij^2 z_ij. And the 'fully linear' McCormick model in (23a-b) is not linear: it has products of W_Fr, W_To, and z. These might be typos, but they are in the defining equations, so the model is mis-specified in print. Also, no code or data is released, which makes independent checking harder.\n\nThe reference list and baselines look fine; the PGLib cases and the DC model from [6] are appropriate. Self-citation is not an issue here, since the prior results are real stepping stones.\n\nWho should read this: anyone working on topology optimization or SOC power-flow relaxations. It is a practical, modest paper. I would send it out for review, but I would want the authors to fix the typos, release code/data, and add an AC redispatch check or at least a sensitivity analysis on a couple of cases.","headline":"An exact single-cone reformulation that is genuinely useful and a plausible negative result, but the AC-deliverability claim is not fully supported and two defining equations appear mis-specified.","tokens_in":12673,"tokens_out":2535,"would_cite":true,"duration_ms":24523,"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":"This paper shows that an exact reformulation of the voltage-cone constraint in the second-order-cone optimal power shutoff problem cuts solve time from about 13,500 seconds to about 46 seconds on the 57-bus case, while fully linear…","keywords":["optimal power shutoff","second-order cone programming","topology optimization","power flow relaxation","wildfire risk","linear relaxation","voltage cone","load delivery"],"falsifier":"Take the binary line decisions produced by the McCormick and secant formulations on a case like case14 or case57 and evaluate them with a full AC power flow solver instead of the SOC redispatch used in the paper. If the AC-feasible delivered load is still far below the linear solutions' estimates, the overestimation claim is confirmed; if AC delivers substantially more than the SOC redispatch indicates, then the SOC proxy, not the linear relaxation, is the source of the measured gap.","tokens_in":11662,"feed_emoji":"⚡","tokens_out":7597,"duration_ms":69027,"temperature":0.7,"pith_summary":"This paper addresses the optimal power shutoff problem, where a grid operator decides which lines to de-energize to balance wildfire risk against load delivery after large-scale outages. The authors try to make the second-order-cone (SOC) power-flow formulation fast enough to be practical by replacing its nonlinear voltage constraints with a small number of linear cuts. Their first result is a simple exact reformulation of the voltage cone, $(W^R_{ij})^2+(W^I_{ij})^2 \\le W^{Fr}_{ij}W^{To}_{ij}$, that both shrinks the formulation and tightens the continuous relaxation, cutting solve time on the 57-bus case from about 13,500 seconds to about 46 seconds. Their second result is negative: the fully linear relaxations they construct overestimate how much load can be delivered, and after re-solving with the SOC power flow they deliver no better load than the much faster DC model. The paper concludes that a good speed-quality trade-off for a fully linear formulation was not found.","feed_headline":"Voltage-cone rewrite cuts outage-plan solve time from 13,500 s to 46 s","feed_subtitle":"Exact voltage-cone rewrite tightens the relaxation; linear cuts still overestimate deliverable load.","key_machinery":"The load-bearing object is the rotated second-order cone inequality $(W^R_{ij})^2+(W^I_{ij})^2 \\le W^{Fr}_{ij}W^{To}_{ij}$, labeled Eq. (17). It replaces the earlier product relaxation that used $W_{ii}W_{jj}z_{ij}$, which required three cone constraints and a binary product. The reformulation works because the per-line squared-voltage variables $W^{Fr}_{ij}$ and $W^{To}_{ij}$ are defined as $W_{ii}z_{ij}$ and $W_{jj}z_{ij}$, so when the line is de-energized the right-hand side is zero and the cone is vacuous, and when the line is energized it equals $W_{ii}W_{jj}$. The paper then builds two linear relaxations around this cone: one using McCormick envelopes on the bilinear product $W^{Fr}_{ij}W^{To}_{ij}$, and one using a secant upper bound; both keep tangent cuts for the quadratic terms but relax the right-hand side, and it is that right-hand-side relaxation that the paper identifies as the source of overestimation.","core_discovery":"The central claim is that the SOC voltage cone in topology-switching power flow can be written exactly as $(W^R_{ij})^2+(W^I_{ij})^2 \\le W^{Fr}_{ij}W^{To}_{ij}$ instead of as a product relaxation that needs multiple cone constraints and an explicit binary multiplier. Because $W^{Fr}_{ij}=W_{ii}z_{ij}$ and $W^{To}_{ij}=W_{jj}z_{ij}$, multiplying the original right-hand side by $z_{ij}^2=z_{ij}$ makes the product form exact; the per-line voltage variables already encode the switching state, so the binary product disappears from the constraint. This reformulation, called SOC-OPS-P, is both smaller and tighter and is the paper's main positive result. The paper further claims that replacing the convex cone with tangent-based linear cuts for thermal limits and either McCormick or secant bounds for the voltage product yields fast but optimistic solutions: the linear models report more deliverable load than the SOC redispatch actually achieves, and their final redispatch performance is no better than DC power flow.","pith_inferences":["If Eq. (17) is used inside restoration or topology-control formulations that also switch lines, the same exactness and tightening should carry over, because the argument only uses the definition of the per-line voltage variables.","The paper's diagnosis points to a concrete next step it does not test: a piecewise-linear outer approximation of the bilinear term $W^{Fr}_{ij}W^{To}_{ij}$ might recover quality without full binaries, at the cost of more constraints.","In speed-critical settings, the results suggest starting from the DC solution and repairing it with an SOC or AC check, rather than solving a large linearized conic model; the paper notes this direction but does not develop it.","Because the evaluation uses SOC redispatch rather than full AC redispatch, the quantitative redispatch ratios are upper bounds on what the paper can conclude about true AC load delivery; the qualitative ranking of formulations is likely more robust than the exact percentages."],"forward_implications":["Any topology-switching SOC power flow model can adopt Eq. (17) to remove the binary product and tighten the relaxation, so the speedup is not limited to the optimal power shutoff problem.","Fully linear relaxations of the voltage cone are not a drop-in replacement for the SOC formulation: they report overly optimistic load delivery and, after redispatch, match rather than beat DC power flow quality.","The secant relaxation is tighter and gives better redispatch than the McCormick relaxation, especially on the worst-case 14-bus network, so it is the better starting point if linear relaxations are pursued further.","Adding more linearization points for the quadratic terms does not improve solution quality; the weak point is the right-hand-side product $W^{Fr}_{ij}W^{To}_{ij}$, not the tangent cuts.","The SOC-OPS-T model, which linearizes only thermal limits, preserves near-identical load delivery with no speedup, suggesting that future gains must come from a better voltage-cone relaxation rather than from thermal-limit cuts."],"supporting_citations":[{"why":"Supplies the topology-switching SOC power flow model on which the OPS formulation is built.","marker":"[4]"},{"why":"Provides the SOC-OPS baseline, the SOC-Redispatch procedure used to evaluate solution quality, and the DC-OPS comparison.","marker":"[6]"},{"why":"Defines the OPS optimization problem and the wildfire-risk objective, including the base implementation used for all variants.","marker":"[3]"},{"why":"Provides the standard test networks, from 5 to 118 buses, on which all formulations are benchmarked.","marker":"[22]"},{"why":"The mixed-integer solver used to run all formulations; the reported solve times depend on its performance.","marker":"[23]"}],"fun_headline_variants":["Exact voltage-cone rewrite cuts outage solve time from 13,500s to 46s","Voltage-cone reformulation makes outage planning 293x faster","SOC-OPS-P exact cone formulation slashes restoration solve time","New cone formulation speeds post-disaster power flow optimization","Linear power-flow cuts overestimate deliverable load in outages"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"It assumes that a solution that passes the relaxed second-order-cone power-flow check will really be able to deliver that much power on the actual AC grid; if the relaxation is loose in these large-outage topologies, the reported load-delivery numbers could be wrong.","fun_headline_variants_meta":{"raw":{"variants":["Exact voltage-cone rewrite cuts outage solve time from 13,500s to 46s","Voltage-cone reformulation makes outage planning 293x faster","SOC-OPS-P exact cone formulation slashes restoration solve time","New cone formulation speeds post-disaster power flow optimization","Linear power-flow cuts overestimate deliverable load in outages"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000545,"raw_usage":{"total_tokens":2624,"prompt_tokens":981,"completion_tokens":1643,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":597,"completion_tokens_details":{"reasoning_tokens":1551}},"tokens_in":597,"tokens_out":1643,"duration_ms":11524,"temperature":1.0,"reasoning_tokens":1551,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:28:13.936980+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the binary line decisions produced by the McCormick and secant formulations on a case like case14 or case57 and evaluate them with a full AC power flow solver instead of the SOC redispatch used in the paper. If the AC-feasible delivered load is still far below the linear solutions' estimates, the overestimation claim is confirmed; if AC delivers substantially more than the SOC redispatch indicates, then the SOC proxy, not the linear relaxation, is the source of the measured gap.","supporting_citations":[{"cited_title":"Relax- ations of ac maximal load delivery for severe contingency analysis,","cited_arxiv_id":null,"evidence_quote":"Supplies the topology-switching SOC power flow model on which the OPS formulation is built."},{"cited_title":"Long solution times or low solution quality: On trade-offs in choosing a power flow formulation for the optimal power shutoff problem,","cited_arxiv_id":null,"evidence_quote":"Provides the SOC-OPS baseline, the SOC-Redispatch procedure used to evaluate solution quality, and the DC-OPS comparison."},{"cited_title":"Balancing wildfire risk and power outages through optimized power shut-offs,","cited_arxiv_id":null,"evidence_quote":"Defines the OPS optimization problem and the wildfire-risk objective, including the base implementation used for all variants."},{"cited_title":"The ieee pes task force on benchmarks for validation of emerging power system algorithms,","cited_arxiv_id":null,"evidence_quote":"Provides the standard test networks, from 5 to 118 buses, on which all formulations are benchmarked."},{"cited_title":"Gurobi optimizer reference manual,","cited_arxiv_id":null,"evidence_quote":"The mixed-integer solver used to run all formulations; the reported solve times depend on its performance."}],"review_version":1}