{"id":"ef7802f8-55ee-4191-84f0-3d484ad53168","arxiv_id":"1908.02832","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A multi-period OPF for multi-frequency HVac systems with back-to-back converters is formulated and solved, reducing simulated peak losses from 4.86% to 1.84% on a modified IEEE 57-bus system.","lead":"This paper formulates an optimal power flow (OPF) for power grids that mix conventional 50/60-Hz AC and low-frequency AC connected by back-to-back converters, and solves it with an interior-point method. If correct, it gives operators a way to schedule generators, capacitors, and converters to reduce transmission losses in low-frequency AC systems.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline 4.86-to-1.84% loss reduction depends on converter switching-loss coefficients (Eqs. 14-16, Table I) that are assumed, mode-ambiguous, and identical across converters of different ratings; a sensitivity test is needed.","rationale":"After reading the paper, the central claim is the demonstration that optimal dispatch reduces losses from 4.86% (Case 1) to 1.84% (Case 3) at peak and removes voltage violations. Several assumptions are fragile, but the converter loss model is the most load-bearing because it directly shapes the objective and the optimal converter dispatch. In contrast, the arbitrary Case 1 baseline is explicitly labeled and a second comparison (Case 2 vs Case 3) does not rely on it; the Hessian and solver claims, while possibly flawed, are cross-checked against IPOPT/BONMIN. The loss model, however, has no independent validation in the paper. The specific weaknesses are: (a) coefficients copied from other converters; (b) no scaling to the 300/200 MVA ratings; (c) Eq. (16) uses one a2 despite Table I giving two modes; and (d) the results are not reproducible because the case-study data and code are not provided. A sensitivity analysis, as described, would settle whether the central numbers are robust. This supports the reader's CONDITIONAL verdict; I would not move to ACCEPT or REJECT on this basis alone.","tokens_in":16406,"tokens_out":18125,"duration_ms":188073,"concrete_test":"Re-run the Case 2 and Case 3 simulations under three variants: (i) multiply all Table I coefficients by 0.5, (ii) multiply by 2.0, and (iii) assign the rectifier a2 to the VSC side that operates as rectifier and the inverter a2 to the inverter side according to the power flow direction, with coefficients scaled by converter rating (e.g., a0, a1, a2 proportional to S_rated/300 MVA). Compare the peak-loss percentages and the ordering of Cases 1-3. If Case 3 no longer beats Case 2 or the peak loss moves by more than, say, 0.5 percentage points, the headline claim is an artifact of the assumed loss model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central loss-reduction numbers are outputs of an optimization whose objective is total generation; converter losses enter through the power balance Eq. (15)-(16). The switching-loss polynomial coefficients a0, a1, a2 (Table I) are taken from prior VSC-HVDC literature, assumed identical for all five converters, and the two converter rating groups (300 MVA vs 200 MVA) are not distinguished. Moreover, Table I lists different a2 for rectifier and inverter modes, yet Eq. (16) uses a single a2 for both VSC sides without specifying which mode applies; the power direction in Table III implies VSC1 and VSC2 operate in different modes, so the model as written is internally ambiguous. Because the optimizer can trade off line losses against converter losses, an inaccurate loss curve can change the optimal dispatch and the reported 1.84% peak loss, or even the ranking of Case 2 (2.59%) and Case 3 (1.84%). The reader's weakest-assumption flag is therefore the most load-bearing element: without a sensitivity check, the headline quantitative claim is not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a multi-period optimal power flow (OPF) for a multi-frequency HVac system in which a conventional 50/60-Hz grid and a low-frequency HVac grid are interconnected by back-to-back VSC converters. The decision variables are generator dispatches, shunt-capacitor steps, and converter active/reactive set points; the objective minimizes weighted generation (equivalently losses) plus a capacitor-switching penalty, subject to ac network constraints, converter loss balance, and converter capability limits. The resulting mixed-integer nonlinear program is solved by a predictor-corrector primal-dual interior-point method with a penalty-based rounding scheme for the discrete capacitor variables. On a 57-bus/8-bus multi-frequency test system, the paper reports peak-load losses of 4.86% under an arbitrary dispatch, 2.59% with OPF on generators and capacitors, and 1.84% when converters are also optimized, with voltage violations eliminated. The solver is compared against IPOPT and BONMIN on two test systems.","tokens_in":16635,"tokens_out":17235,"duration_ms":167197,"significance":"If the quantitative claims hold, the paper provides a useful, clearly specified OPF model for an emerging transmission technology, and the solution method is a reasonable extension of PCPDIPM to MINLP with discrete shunt elements. The use of rectangular coordinates with precomputed constant Hessians for network constraints is a practical efficiency idea, and the comparison with IPOPT and BONMIN gives external grounding for the solution quality. The main value is the complete formulation (converter losses, capability curves, multi-period capacitor switching) rather than a fundamentally new algorithm. However, the headline loss-reduction numbers depend on assumed converter loss coefficients and on the correctness of the published Hessian formulas, so the quantitative claims are not yet fully established.","major_comments":[{"comment":"Table I lists different a2 values for rectifier (4.400×10^-3) and inverter (6.667×10^-3) modes, but Eq. (16) uses a single a2 for both VSC1 and VSC2 sides of the back-to-back converter. In the operating point of Table III, P_conv_s is negative and P_conv_l is positive for every converter, so VSC1 and VSC2 operate in different modes; applying one a2 is therefore internally ambiguous. Because converter losses appear in the power-balance constraint that couples the two sides, this ambiguity can change the optimal converter dispatch and the reported 1.84% peak loss. Please introduce separate a2,rect and a2,inv (or state explicitly that the same value is used for both modes and justify it).","section":"§IV-E, Eq. (16), Table I"},{"comment":"The converter loss coefficients in Table I are assumed identical for all five converters, even though Converters A and B have Srated = 300 MVA while C-E have Srated = 200 MVA, and the coefficients are taken from previous VSC-HVDC literature rather than measured or fitted for these converters. Since the OPF can trade off line losses against converter losses, the optimal dispatch—and hence the 4.86% to 1.84% reduction and the Case 2 vs Case 3 ranking—is sensitive to this assumed loss curve. A sensitivity study over a0, a1, and a2 (e.g., ±20% and mode-dependent variation) is needed to establish the robustness of the headline result; without it, the central quantitative claim is not supported at the reported precision.","section":"§VI-B, Figs. 7-8"},{"comment":"The Appendix contains algebraic errors in the claimed exact Hessian matrices. For constraint (16), the a1 terms should scale as |S| (for voltage derivatives) and as |S|^-3 (for P/Q derivatives), but the printed formulas use |S|^2 and |S|^-6; e.g., ∂2g/∂P^2 should be 2(R+a2)/V^2 + a1 Q^2/(V |S|^3), not with |S|^6 in the denominator. For constraint (18), the voltage second derivative has the wrong sign: hIconv = P^2+Q^2-(Imax)^2(e^2+f^2), so ∂2h/∂e^2 = -2(Imax)^2, not +2(Imax)^2. For constraint (20), the e-derivative should have a b-term Q+b(3e^2+f^2) and a separate -2k_V^2 term, not the printed Q+b(e^2+3f^2)-2k_V^2 inside the 4b factor; the f-derivative has a similar swap, and constraint (21), which has opposite signs on the g and b terms, requires its own formulas that are not given. Because the exact Hessians are a stated contribution and are used to justify the solver's efficiency, these errors must be corrected.","section":"Appendix"},{"comment":"The penalty-based rounding of shunt capacitors to discrete values is introduced without a feasibility or optimality guarantee. The comparison with BONMIN in Table V covers only two test systems and reports only objective values and iteration counts; it does not demonstrate that the final discrete solution satisfies all constraints or that the penalty heuristic reaches the same feasible set as BONMIN. Please either provide a formal statement of the conditions under which the rounding step preserves feasibility/optimality, or explicitly label the method as heuristic and add a post-hoc constraint-violation check for the reported solutions.","section":"§V, Table V"}],"minor_comments":[{"comment":"The subscripts i and k are mixed in Eq. (16) (P_conv_s,i, R1,i) for the same converter; use k consistently.","section":"§IV-E, Eq. (16)"},{"comment":"Case 1 is described only as 'a given power dispatch'; please specify the generator, shunt, and converter set points so the 4.86% baseline is reproducible.","section":"§VI-B"},{"comment":"Table V would be more informative if it reported feasibility tolerances and the discrete feasibility of the IPOPT solution, since IPOPT solves the continuous relaxation only.","section":"§VI-C, Table V"},{"comment":"The conclusion's '3% loss reduction' should be stated as '3.02 percentage points' (4.86% to 1.84%) to avoid misinterpretation.","section":"§VII"},{"comment":"The text mentions 'red numbers' in Table III, but no red numbers are visible in the printed table; clarify which entries are binding.","section":"§VI-B, Table III"},{"comment":"The choice α2 = 0.2 in Case 3 is not justified; a brief sensitivity discussion would help the reader understand the loss-versus-switching trade-off.","section":"§VI-B"},{"comment":"The claim that 'no research has been done to solve OPF in multi-frequency HVac power systems' is difficult to verify; consider softening to 'to the best of our knowledge' and citing adjacent HVDC/multi-frequency OPF works.","section":"§II"}],"recommendation":"major_revision","confidential_remarks":"The paper is an incremental but solid extension of the authors' prior multi-frequency power-flow work [10]. The two blocking issues—Hessian appendix errors and converter-loss ambiguity—are fixable in revision. The single test system and arbitrary baseline limit the strength of the numerical claims, but do not by themselves warrant rejection. No concerns about author conduct or citation behavior."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is the first OPF formulation for multi-frequency HVac systems with back-to-back converters, and the exact Hessian work is real. The central loss-reduction numbers rest on a converter loss model that has an internal inconsistency, so treat the 4.86-to-1.84% as indicative, not established.\n\nWhat's new: the paper extends the authors' earlier power-flow tool to multi-period OPF with converter dispatch. The rectangular-coordinate form gives constant Hessians for network constraints, and they derive exact second derivatives for the converter power balance and capability constraints. That is genuinely useful and not in the earlier literature. They also cross-check against IPOPT and BONMIN on the same systems; the objective values match closely, which supports the solver's correctness. The discrete capacitor handling is heuristic but standard, and it converges to discrete values in the reported runs.\n\nSoft spots: the converter switching-loss model is the load-bearing input. Table I gives different a2 for rectifier and inverter modes, but Eq. (16) applies a single a2 to both VSC sides. The power directions in Table III imply VSC1 operates as a rectifier and VSC2 as an inverter, so the model as written is internally ambiguous. Since the objective is loss minimization, the optimizer can shift power between line losses and converter losses, so the reported 1.84% peak loss could move if the correct a2 were used. A sensitivity test over a2 and a1 would settle this. The case study is one modified 57-bus system with an arbitrary baseline for Case 1; that is fine for a proof-of-concept but not a strong empirical claim. No code or complete data is released, so the exact results are not independently reproducible, though the solver comparison is credible.\n\nOverall, the formulation and solver are a genuine contribution for a niche but emerging transmission option. The loss-reduction headline is plausible but should be read as conditional on the loss model. I would send it to peer review; a good referee would ask for a mode-consistent loss model and a sensitivity analysis.","headline":"First multi-frequency HVac OPF with real exact-Hessian content; headline loss numbers hinge on an internally inconsistent converter loss model.","tokens_in":17138,"tokens_out":1739,"would_cite":true,"duration_ms":18890,"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 claims that a multi-period optimal power flow dispatching generators, shunt capacitors, and back-to-back converters together cuts peak-load losses in a multi-frequency HVac system from 4.86% to 1.84% and eliminates voltage…","keywords":["multi-frequency power systems","low-frequency AC transmission","optimal power flow","back-to-back converters","converter loss model","interior-point method","loss minimization","voltage regulation"],"falsifier":"Run the same multi-period OPF with switching-loss coefficients measured from the actual converters instead of the assumed values in Table I; if the optimal dispatch or the 1.84% peak-loss figure changes materially, the claimed loss reduction rests on the unmeasured assumption.","tokens_in":16204,"feed_emoji":"⚡","tokens_out":14168,"duration_ms":116989,"temperature":0.7,"pith_summary":"This paper tries to establish that the emerging low-frequency high-voltage ac (LF-HVac) transmission concept can be operated with substantially lower losses by solving a multi-period optimal power flow (OPF) that treats back-to-back converters, generators, and shunt capacitors as coordinated control resources. On a multi-frequency system built from a standard 57-bus test network, the proposed dispatch lowers peak-load losses from 4.86% to 1.84% and keeps every load-bus voltage inside its limit across a 24-hour load profile. If true, this means LF-HVac corridors, which need frequency conversion to connect to 50/60-Hz grids, can deliver their advertised low-loss advantage only when the converters' power injections are optimized rather than chosen arbitrarily. The paper's contribution is a tractable formulation and solver for that optimization: a mixed-integer nonlinear program in rectangular coordinates, solved with a modified interior-point method that handles discrete capacitor steps and exact Hessians.","feed_headline":"Cuts multi-frequency grid losses from 4.86% to 1.84%","feed_subtitle":"A multi-period OPF coordinates generators, capacitors, and converters to cut losses and clear voltage violations.","key_machinery":"The load-bearing mechanism is the multi-period OPF in rectangular coordinates: bus voltages are split into real and imaginary parts so the Hessian matrices of the nodal power-balance equalities become constant and can be precomputed once in compressed sparse-row storage, while discrete capacitor dispatches are forced to their nearest allowed values by a quadratic penalty inside a predictor-corrector primal-dual interior-point method. A back-to-back converter station is modeled as two voltage-source converters sharing a dc link, with Joule losses in transformers and phase reactors plus a switching-loss polynomial $a_0 + a_1 I + a_2 I^2$; that loss model enters the OPF as a power-balance equality whose exact Jacobian and Hessian are derived in the appendix. This combination makes a nonconvex mixed-integer nonlinear program solvable in about one second per time step on the test system.","core_discovery":"The central claim is that optimal dispatch of generators, shunt capacitors, and back-to-back converters in a multi-frequency HVac transmission system is a solvable, multi-period MINLP whose solution substantially reduces losses. In the tested system, the optimizer cuts losses from 4.86% to 1.84% at peak load and eliminates voltage violations throughout a simulated day, while also reducing capacitor switching operations when that is penalized. The paper argues the key enabler is writing the OPF in rectangular coordinates, where the Hessians of nodal power balance constraints are constant and can be precomputed in compressed sparse-row form, and deriving the exact Hessians of the converter power-balance and capability constraints. The resulting predictor-corrector interior-point framework converges to discrete capacitor settings and matches the solution quality of a general mixed-integer solver with fewer iterations.","pith_inferences":["If the loss model is accurate, the same OPF could be used in planning: the marginal loss value at each converter site would rank candidate LF-HVac corridors by energy savings.","The reported 1.84% peak-loss figure assumes the switching-loss coefficients in Table I are identical across converters; a sensitivity sweep over plausible coefficient ranges would reveal how much of the reduction is an artifact of that assumption.","Because the formulation treats the LF-HVac grid as load-free, extending it to serve loads inside the low-frequency grid could change both the optimum and the converter dispatch, a natural next step not explored in the paper.","The same warm-started interior-point machinery could be adapted to a rolling-horizon online dispatch, using the previous time step's solution to track load changes faster than the reported one-second solve."],"forward_implications":["In a multi-frequency HVac system, back-to-back converters are not just frequency couplers; their active and reactive dispatch can be co-optimized with generators and capacitors to cut system losses.","The proposed OPF keeps load-bus voltages within limits across the daily load profile, eliminating the overvoltages observed under arbitrary dispatch.","Penalizing capacitor switching in the objective reduces both switching operations and losses, meaning converter dispatch can substitute for capacitor-bank action.","The rectangular-coordinate formulation with precomputed constant Hessians makes each time step fast enough (about one second) for operational use on systems of this size.","The exact Hessian expressions for converter constraints can be reused in OPF formulations for hybrid HVac-HVdc systems, as the appendix notes."],"supporting_citations":[{"why":"Supplies the multi-frequency power-flow formulation and solver used for warm starts and the no-OPF base case.","marker":"[10]"},{"why":"Provides the generalized steady-state VSC model and converter loss expressions (Joule plus switching) used in the OPF.","marker":"[11]"},{"why":"Foundational interior-point method in rectangular coordinates that the solution framework extends.","marker":"[17]"},{"why":"Supplies the discrete-variable penalty technique that forces capacitor dispatch to practical values.","marker":"[18]"},{"why":"An interior-point NLP solver used as a comparison baseline in the convergence study.","marker":"[20]"},{"why":"A mixed-integer nonlinear solver used as a comparison baseline for discrete convergence and iteration count.","marker":"[22]"},{"why":"Source of the converter current and voltage capability constraints embedded in the OPF.","marker":"[25]"},{"why":"The 57-bus test system the case study modifies to build the multi-frequency network.","marker":"[28]"}],"fun_headline_variants":["OPF cuts multi-frequency HVac losses from 4.86% to 1.84%","Multi-period OPF coordinates dispatch to cut multi-frequency grid losses","Predictor-corrector OPF cuts multi-frequency losses from 4.86% to 1.84%","Optimal dispatch in multi-frequency grids reduces losses to 1.84%","Coordinated dispatch of generators, capacitors, converters cuts HVac losses"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Every back-to-back converter in the study is assumed to have the same switching-loss polynomial, with coefficients taken from prior references rather than measured for the actual hardware, and both the loss objective and the converter dispatch depend on that model.","fun_headline_variants_meta":{"raw":{"variants":["OPF cuts multi-frequency HVac losses from 4.86% to 1.84%","Multi-period OPF coordinates dispatch to cut multi-frequency grid losses","Predictor-corrector OPF cuts multi-frequency losses from 4.86% to 1.84%","Optimal dispatch in multi-frequency grids reduces losses to 1.84%","Coordinated dispatch of generators, capacitors, converters cuts HVac losses"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00142,"raw_usage":{"total_tokens":5704,"prompt_tokens":888,"completion_tokens":4816,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":4707}},"tokens_in":504,"tokens_out":4816,"duration_ms":39851,"temperature":1.0,"reasoning_tokens":4707,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:32:06.275649+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same multi-period OPF with switching-loss coefficients measured from the actual converters instead of the assumed values in Table I; if the optimal dispatch or the 1.84% peak-loss figure changes materially, the claimed loss reduction rests on the unmeasured assumption.","supporting_citations":[{"cited_title":"Power flow in a multi- frequency hvac and hvdc system: Formulation, solution, and validation,","cited_arxiv_id":null,"evidence_quote":"Supplies the multi-frequency power-flow formulation and solver used for warm starts and the no-OPF base case."},{"cited_title":"Generalized steady-state VSC MTDC model for sequential AC/DC power flow algorithms,","cited_arxiv_id":null,"evidence_quote":"Provides the generalized steady-state VSC model and converter loss expressions (Joule plus switching) used in the OPF."},{"cited_title":"On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming,","cited_arxiv_id":null,"evidence_quote":"An interior-point NLP solver used as a comparison baseline in the convergence study."},{"cited_title":"An algorithmic framework for convex mixed integer nonlinear programs,","cited_arxiv_id":null,"evidence_quote":"A mixed-integer nonlinear solver used as a comparison baseline for discrete convergence and iteration count."},{"cited_title":"[Online]","cited_arxiv_id":null,"evidence_quote":"The 57-bus test system the case study modifies to build the multi-frequency network."}],"review_version":1}