{"id":"1599e16f-9595-4e3c-af6c-86d623974154","arxiv_id":"2501.18732","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A bilevel optimization shows that a single zero-price quantity bid by renewable producers can replicate the system-optimal outcome of any multi-price bidding curve in two-settlement markets, and it cuts simulated NYISO costs by 36%.","lead":"The paper develops an optimization framework for how wind and solar plants should bid in day-ahead electricity markets, accounting for expensive corrections in the real-time market. On a large New York grid model, it reports 36% lower hourly system costs than bidding expected output, approaching the theoretical ideal.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported 36% savings may rest on an unchecked McCormick relaxation gap: the LP solved in §5.1 is a relaxation of the true bilevel problem, so S_BiD could be a lower bound rather than an achievable system cost.","rationale":"The reader's weakest assumption is that the proof of Theorem 1 and the solution method require convexity from the binary-UC relaxation. While the solution method does require an LP, the proof of Theorem 1 in Appendix A is a set-based constructive argument: it shows that the total VRE dispatch from any optimal multi-segment solution can be collapsed into a single zero-price segment without changing the day-ahead conventional dispatch or the real-time feasible set. This argument does not use KKT or strong duality, and it remains valid even if binary commitment variables were present, because feasibility and the cost comparison are preserved when total VRE output is held fixed. Thus the convexity assumption is not the most load-bearing weak point for the theoretical claim. The more direct threat to the central claim is the numerical solution method: the McCormick-envelope LP solved in §5.1 is a relaxation, so the reported S_BiD could be an optimistic lower bound. The paper does not quantify the gap or report a simulation-based verification of the optimized bids, which is essential before the 36% reduction can be accepted. No code or data are provided, but the proposed test is straightforward to run on the existing case-study data. The verdict should remain conditional until this gap is characterized.","tokens_in":14282,"tokens_out":15458,"duration_ms":147286,"concrete_test":"Take the LP-derived bid curves W* from the single- and multi-segment experiments, re-solve the day-ahead LP (1) with those bids, then solve the real-time problem (3) for each scenario, and compute the resulting expected system cost. Compare this simulated cost with the reported S_BiD. If the simulated cost exceeds $275k by more than a small tolerance (e.g., 1–2%), the 36% savings claim is inflated and the McCormick gap is material. As a second check, on a small instance (e.g., the 24-bus system used in Morales et al. 2014) solve the exact bilevel problem via Big-M KKT reformulation and compare its optimal value to both the LP relaxation bound and the simulated cost.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's numerical claim of a 36% cost reduction (S_BiD = $275k vs. S_MyD = $430k) is computed with the solution method in §5.1, which replaces the lower-level KKT conditions with strong duality and then relaxes the resulting bilinear terms λ^W·W using McCormick envelopes. This produces an LP that is an outer relaxation of the true bilevel problem, so its optimal value is a lower bound on the true expected system cost. If the reported $275k is the relaxation objective rather than the cost obtained by simulating the optimized bid curves back into the day-ahead and real-time market models, the savings may be overstated. The paper does not report the relaxation gap, nor does it state whether S_BiD was computed by re-solving the lower-level LP with the optimized W* and then evaluating RTM costs. Without such validation, the 36% figure is not verified as an achievable cost under the market-clearing rules. Note that Theorem 1 itself is proved in Appendix A using a constructive feasibility argument that does not invoke KKT or strong duality; it relies only on the linearity of costs and the 'cheapest-segments-first' property, so the binary-UC/convexity issue raised by the reader mainly affects the solution algorithm, not the theorem's internal validity. The load-bearing gap is therefore the unquantified tightness of the McCormick relaxation that underpins the numerical headline.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a bilevel optimization framework for designing day-ahead bidding curves for variable renewable energy (VRE) in a two-settlement electricity market. The upper level chooses VRE bidding prices and quantities; the lower level is the day-ahead market-clearing linear program; the objective is the true day-ahead cost plus the expected real-time redispatch cost. The central theoretical result (Theorem 1) states that, when VRE marginal cost is zero, a single-segment zero-price bid with an optimized quantity (Problem BiD-q) attains the same expected system cost as the general multi-price multi-segment problem. Numerical experiments on a 1576-bus NYISO model with relaxed unit commitment report S_MyD = $430k, S_BiD = $275k, and S_StD = $263k, corresponding to a 36% hourly system-cost reduction relative to myopic forecast bidding.","tokens_in":14605,"tokens_out":7105,"duration_ms":73629,"significance":"The constructive proof of Theorem 1 in Appendix A is a genuine structural insight: it does not assume its conclusion, it does not rely on KKT or strong duality, and it is coherent within the stated convex LP market model. If the numerical results were verified as achievable costs, the paper would offer a practical, scalable benchmark for VRE bidding in sequential deterministic markets and a clear comparison with the stochastic-dispatch ideal. The main caveats are that the reported S_BiD comes from a McCormick-envelope relaxation whose gap is not reported, and that all numerical claims depend on the relaxation of binary unit commitment. These caveats are load-bearing for the 36% headline, so the significance is conditional until the gap is quantified or the optimized curves are evaluated in the original market-clearing problems.","major_comments":[{"comment":"The solution method in §5.1 replaces the lower-level KKT conditions with strong duality and then relaxes the bilinear terms λ^W·W using McCormick envelopes. This produces an LP that is an outer relaxation of the true bilevel problem, so its optimal value is a lower bound rather than an achievable expected system cost. The paper does not state whether the reported S_BiD = $275k (and the BiD cost curves in Figs. 3–5) is the relaxation objective or the cost obtained by re-clearing the original day-ahead and real-time market problems with the optimized W*. Without this information, the 36% cost reduction is not verified as an achievable market outcome. Please either report the relaxation gap (for example, by evaluating the optimized bid curve in the original lower-level problems, or by quantifying the gap on a smaller instance where the exact bilevel problem can be solved) or explicitly label the reported S_BiD as a lower bound.","section":"§5.1, §5.5"},{"comment":"The market-clearing models relax binary unit-commitment decisions to continuous variables 'to preserve convexity'. All numerical results, including the S_StD benchmark and the displayed ordering S_MyD ≥ S_BiD ≥ S_StD in §5.3, are therefore for a convexified unit-commitment model. Actual day-ahead markets solve a mixed-integer program, so the 36% savings should be presented as conditional on this relaxation, and the paper should state explicitly that the proposed LP solution method does not directly apply to a binary-UC market-clearing problem. This does not invalidate Theorem 1, but it is essential context for the numerical claims.","section":"§2, after Eq. (1); §5"},{"comment":"The inequality S_MyD ≥ S_BiD ≥ S_StD is asserted with a citation to prior work. If S_BiD is taken from the McCormick relaxation, the inequality need not hold for the true bilevel optimal value; the theoretical ordering should be stated for the exact Problem BiD, and the numerical comparison should distinguish the relaxation lower bound from a feasible system cost. This distinction is necessary to support the paper's 'absolute dominance' claim.","section":"§5.3"}],"minor_comments":[{"comment":"The heading 'Single-Segment Biding Curves' contains a typo; it should read 'Bidding'.","section":"§5.4"},{"comment":"In the proof after Eq. (A.3), the indices kα and tα are introduced but Eq. (A.4) writes pW‡_{k,t} without subscripts; the notation should be made consistent for readability.","section":"Appendix A"},{"comment":"The footnote says that multi- and single-segment curves theoretically achieve the same cost but simulation results differ slightly; this difference is likely due to the LP relaxation and should be stated as such rather than treated as pure numerical noise.","section":"§5.5, footnote 3"},{"comment":"The solution method is delegated entirely to the authors' prior work (Zhao et al., 2024); a journal paper should include at least the key McCormick envelope constraints for λ^W·W or a precise equation-level reference so that the relaxation is reproducible.","section":"§5.1"},{"comment":"The scenario-generation procedure via PGscen and the choice of 20 scenarios are not described beyond a citation; a few sentences on scenario quality, sample size sensitivity, or the scaling of wind capacity would help the reader assess the robustness of the reported cost savings.","section":"§5.2"}],"recommendation":"major_revision","confidential_remarks":"The core theoretical contribution is credible and likely publishable as a market-design insight: the proof of Theorem 1 is constructive and does not depend on the numerical approximation. The main risk is the numerical verification: the advertised 36% reduction rests on a McCormick-envelope relaxation whose gap is not quantified, and the paper does not state whether the reported S_BiD was obtained by evaluating the optimized bid curves in the original market-clearing problems. I would recommend requiring either a feasible evaluation of the optimized curves or a quantified relaxation gap before acceptance. The paper would also benefit from a clearer statement that all numerical claims are conditional on relaxed unit commitment, and from making the evaluation pipeline fully specified."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First read: Theorem 1 is the thing to remember. It says that in a two-settlement market with zero-marginal-cost VRE, optimizing any multi-price multi-segment bid curve is no better than optimizing a single zero-price quantity bid. The proof in Appendix A is constructive and doesn't lean on KKT or strong duality; it uses the monotonicity of segment prices and a marginal substitution argument. That's a clean, new result and it makes the paper worth reading even if you never touch the numerics.\n\nThe paper also does something useful in the case study: on a 1576-bus NYISO model, the optimized bid cuts hourly expected cost from $430k (myopic forecast bid) to about $275k, close to the $263k stochastic ideal. The bit about multi-segment curves with fixed prices is a reasonable extension of their earlier McCormick-envelope method, and the figures showing how the optimized curves differ from forecasts are informative.\n\nNow the soft spots. The biggest one is the claim attached to that $275k number. Section 5.1 replaces the lower-level KKT conditions with strong duality and then relaxes the bilinear terms with McCormick envelopes. That yields an LP whose objective is a lower bound on the true bilevel objective. The paper never says explicitly whether S_BiD = $275k comes from re-solving the original DA and RT LPs with the optimized bid quantities, or whether it's just the relaxation objective. If it's the latter, the 36% savings is not demonstrated as achievable. A referee should insist on this point. Second, the relaxation gap is not quantified here; they point to the prior paper, which is fine for the method but not for trusting the headline number. Third, the test is a single 4-hour window (7–10 a.m.) with 20 scenarios; that's a thin basis for 'absolute dominance.' Fourth, the model relaxes binary unit commitment to a convex LP, which is stated up front and is a standard simplification, but it does mean the real-market guarantee is approximate.\n\nThe theorem, again, is solid under the stated assumptions. The circularity worry doesn't land: the stochastic benchmark is an independent lower bound and MyD is a feasible instance of BiD.\n\nVerdict: This deserves a serious referee. The theoretical result is a genuine contribution and the numerical framework is scalable. The referee should ask for a precise description of how S_BiD was evaluated, a gap measure for the McCormick relaxation, and ideally one more case or a longer horizon. None of that kills the paper; it just needs to be tightened.","headline":"Clean theorem, plausible case study, but the 36% headline needs a check on how S_BiD was actually computed.","tokens_in":15142,"tokens_out":4212,"would_cite":true,"duration_ms":38517,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A single zero-price bid segment can match any complex renewable bidding curve.","keywords":["bilevel optimization","bidding curves","variable renewable energy","two-settlement electricity markets","day-ahead market","real-time market","strong duality","McCormick envelope"],"falsifier":"Run the bilevel comparison on a small network where unit commitment is enforced as binary on/off, such as a 6-bus system with one wind farm and three thermal units; if the best multi-segment non-zero-price bidding curve achieves a strictly lower expected system cost than the best single-segment zero-price curve, Theorem 1 is false.","tokens_in":14110,"feed_emoji":"⚡","tokens_out":5881,"duration_ms":49257,"temperature":0.7,"pith_summary":"This paper asks how wind and solar generators should bid in the day-ahead half of a two-settlement electricity market when their real-time output is uncertain and redispatch is costly. It proposes a bilevel optimization that chooses day-ahead bidding curves to minimize the expected sum of day-ahead and real-time system costs, anticipating how the market will clear in each scenario. The central result is a theorem: with zero-marginal-cost renewables, a single-segment curve bidding zero price and an optimized quantity achieves the same expected system cost as any multi-segment, multi-price curve. On a 1576-bus NYISO model, the optimized curves reduce hourly system cost by 36% relative to bidding the expected forecast, and come close to the idealized stochastic-dispatch benchmark. The practical upshot is that operators can coordinate the two settlements through simple quantity benchmarks rather than complex price curves.","feed_headline":"One zero-price bid segment matches any complex renewable bid curve","feed_subtitle":"A single zero-price renewable bid with the right quantity cuts NYISO hourly cost by 36%.","key_machinery":"The central object is the bilevel program Problem BiD: the upper level picks VRE bidding prices and quantities, the lower level is the day-ahead LP market clearing, and the objective adds the expected real-time redispatch cost. The argument runs through Theorem 1's construction, which collapses any optimal multi-segment curve to a single zero-price segment whose quantity is the day-ahead VRE dispatch. Computationally, the paper replaces the lower-level KKT complementarity conditions by strong duality and linearizes the resulting bilinear terms with McCormick envelopes, a relaxation that replaces bilinear products by linear inequalities, yielding an LP relaxation that scales to the 1576-bus NYISO test system.","core_discovery":"The paper states and proves Theorem 1: the general bilevel bidding-curve problem (BiD), which jointly optimizes prices and quantities of multi-segment day-ahead VRE bids, and the quantity-only problem (BiD-q), which uses a single segment at price zero, achieve the same expected system cost. The proof shows that from any optimal multi-segment solution, one can set the single-segment quantity equal to the total VRE quantity dispatched in the day-ahead market; the resulting day-ahead schedule remains optimal for the zero-price problem and yields the same real-time feasible set, hence the same expected cost. A corollary extends this to multi-segment curves as long as one segment is priced at zero. In the numerical case, the optimized bid produces an hourly system cost of $275k versus $430k for the myopic expected-forecast bid and $263k for stochastic dispatch.","pith_inferences":["Because the proof requires the day-ahead clearing to be an LP, applying the same benchmark in a market with binary unit commitment could create a gap between the single-segment zero-price bid and the true optimal multi-segment bid; a dedicated counterexample search on a small mixed-integer system would reveal how large that gap can be.","The theorem suggests a simple coordination instrument: publish only optimal VRE quantities at zero price, which would preserve system cost but change revenue distribution, so fairness and cost-recovery questions would need separate treatment.","The same bilevel machinery could be extended to storage or demand response by adding intertemporal constraints; the theorem would likely fail there because storage has opportunity costs, so multi-segment price signals may regain a role.","One testable prediction is that in systems with high VRE penetration and little fast-ramping capacity, the cost gap between myopic forecast bids and optimized quantity benchmarks should grow as forecast error and ramping scarcity increase."],"forward_implications":["System operators can replace complex VRE price bids with a single zero-price quantity benchmark for each producer and hour without losing expected-cost optimality.","Multi-segment bids remain usable in practice: as long as one segment carries a zero price, the bilevel framework can set segment quantities to approximate the optimum.","On the NYISO 1576-bus system, the framework achieves a 36% hourly cost reduction over myopic expected-forecast bids, approaching the stochastic-dispatch ideal.","The strong-duality and McCormick-envelope LP relaxation is what makes the approach scale to realistic systems, unlike Big-M KKT reformulations tested only on small cases.","The framework can serve as a centralized benchmark to guide or regulate VRE bidding, for example through risk scores."],"supporting_citations":[{"why":"Supplies the strong-duality and McCormick-envelope LP relaxation method that makes the bilevel problem solvable at NYISO scale.","marker":"Zhao et al. (2024)"},{"why":"Introduced the bilevel approach to adjusting day-ahead wind schedules in deterministic energy markets, the baseline this paper generalizes to bidding curves.","marker":"Morales et al. (2014)"},{"why":"Provides the 1576-bus NYISO network dataset used in the case study.","marker":"Greene (2022)"},{"why":"Provides the scenario-generation method used to create the 20 wind-load scenarios for the NYISO case.","marker":"Carmona and Yang (2022)"},{"why":"Supports the simplified convex unit-commitment formulation whose LP structure the optimality proof relies on.","marker":"Kasina et al. (2014)"},{"why":"Supports the convex UC relaxation and the two-settlement market model.","marker":"Kazempour and Hobbs (2017)"},{"why":"Defines the envelope relaxation used to linearize bilinear terms in the KKT and strong-duality reformulation.","marker":"McCormick (1976)"}],"fun_headline_variants":["Zero-price wildcard: one bid segment matches complex renewable curves","Single zero-price bid segment beats myopic forecasting","Renewable bidding: one zero-price segment suffices","Simplest optimal bid: zero price, one segment","Zero-price single segment matches optimal bidding curves"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof and the solution method need both market-clearing problems to be convex linear programs, which the paper forces by relaxing binary unit-commitment decisions to continuous values; real day-ahead markets with discrete unit commitment are outside the guarantee.","fun_headline_variants_meta":{"raw":{"variants":["Zero-price wildcard: one bid segment matches complex renewable curves","Single zero-price bid segment beats myopic forecasting","Renewable bidding: one zero-price segment suffices","Simplest optimal bid: zero price, one segment","Zero-price single segment matches optimal bidding curves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000651,"raw_usage":{"total_tokens":3008,"prompt_tokens":990,"completion_tokens":2018,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":606,"completion_tokens_details":{"reasoning_tokens":1942}},"tokens_in":606,"tokens_out":2018,"duration_ms":12841,"temperature":1.0,"reasoning_tokens":1942,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T22:41:59.133638+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the bilevel comparison on a small network where unit commitment is enforced as binary on/off, such as a 6-bus system with one wind farm and three thermal units; if the best multi-segment non-zero-price bidding curve achieves a strictly lower expected system cost than the best single-segment zero-price curve, Theorem 1 is false.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the strong-duality and McCormick-envelope LP relaxation method that makes the bilevel problem solvable at NYISO scale."},{"cited_title":"M., Zugno, M., Pineda, S., and Pinson, P","cited_arxiv_id":null,"evidence_quote":"Introduced the bilevel approach to adjusting day-ahead wind schedules in deterministic energy markets, the baseline this paper generalizes to bidding curves."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the 1576-bus NYISO network dataset used in the case study."},{"cited_title":"Joint Stochastic Model for Electric Load, Solar and Wind Power at Asset Level and Monte Carlo Scenario GenerationRen\\'e Carmona \\& Xinshuo Yang","cited_arxiv_id":"2209.13497","evidence_quote":"Provides the scenario-generation method used to create the 20 wind-load scenarios for the NYISO case."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supports the simplified convex unit-commitment formulation whose LP structure the optimality proof relies on."},{"cited_title":"and Hobbs, B","cited_arxiv_id":null,"evidence_quote":"Supports the convex UC relaxation and the two-settlement market model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the envelope relaxation used to linearize bilinear terms in the KKT and strong-duality reformulation."}],"review_version":1}