{"id":"60caaf21-fb36-44f9-8f48-65ef1a49743e","arxiv_id":"2506.01426","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":5,"one_line_summary":"A co-design optimization of hybrid storage for truck charging finds the battery+supercapacitor+flywheel mix cheapest by 1.96% over battery-only, yet the advantage over battery+supercapacitor is only 0.001 M€.","lead":"This paper jointly sizes and schedules a truck-charging depot microgrid with batteries, supercapacitors, and flywheels, testing the design on a Dutch distribution center. The full hybrid cuts 20-year total cost by about 2% versus batteries alone, but it is essentially tied with a battery-plus-supercapacitor design.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline result that the fully hybrid design (Exp4) is best is not supported by Table 2: Exp4 and Exp2 total costs differ by 0.001 k€ while the solver gap is 1%, and the text's OpEx/CapEx figures contradict the table.","rationale":"The reader's verdict is REJECT, and my analysis supports that verdict, so no verdict adjustment is needed. However, my primary load-bearing concern differs from the reader's weakest_assumption. The reader focused on the synthetic 30-day month and 60-minute time step as the weakest point. That concern is valid and undermines the reliability of any cost saving, including the 1.96% figure. But an even more direct and decisive flaw is visible within the reported results themselves: Exp2 and Exp4 total costs differ by 0.001 k€, which is far below the 1% solver gap, so the optimizer cannot distinguish these designs. The abstract and text single out Exp4 as 'best,' but the data show a tie with Exp2. Additionally, the prose and Table 2 disagree on OpEx and CapEx for Exp4, indicating that the reported numbers are internally inconsistent. Because the central claim depends on a difference that the solver cannot resolve, the paper's headline result is unsupported even before considering the representativeness of the input data. I therefore agree with the reader's overall rejection, but I would prioritize the numerical-tie and internal-inconsistency issue over the synthetic-data issue as the most load-bearing concern. A concrete test that re-runs Exp2 and Exp4 with a tight MIP gap and checks whether Table 2's numbers are reproducible would settle the matter: if the tie persists or the numbers do not reproduce, the central claim fails unequivocally.","tokens_in":8661,"tokens_out":6838,"duration_ms":71903,"concrete_test":"Re-run the optimization for Experiments 2 and 4 with a MIP gap of 0.01% (or tighter) and with the unit costs and capacities from Table 1, then verify that the reported total costs, CapEx, and OpEx in Table 2 reproduce exactly. If the optimal total costs of Exp2 and Exp4 still differ by less than the solver gap, or if the reported OpEx/CapEx values cannot be reproduced from the model equations, then the central claim that the fully hybrid design is the best is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—that the fully hybrid Battery+Supercapacitors+Flywheel solution achieves the lowest overall cost (1.96% lower than battery-only)—is not supported by the reported data. In Table 2, Experiment 4 (Exp4) has a total cost of 22.386 k€, while Experiment 2 (Battery+Supercapacitors, no flywheel) has 22.387 k€. The difference is 0.001 k€, i.e., one euro on a multi-million-euro cost, while the solver gap is stated as less than 1% (approximately ±220 k€ on 22,000 k€). The optimizer cannot resolve this difference; Exp2 and Exp4 are effectively tied. The abstract's 'best outcomes' claim therefore reduces to numerical noise. Independent of the solver gap, the prose contradicts Table 2: the text says Exp4 'Opex is the lowest (20.287 k€)' but Table 2 lists OpEx 19.757 k€; the text says CapEx is '2.520 k€' while the table lists 2.629 k€. These inconsistencies mean the reported supporting numbers are not internally reproducible. The synthetic-month and hourly-time-step concern raised by the reader is real and compounds this: with τ=60 min, the model cannot represent the sub-hourly transients that justify supercapacitors and flywheels, so even the 1.96% saving over battery-only is not anchored to the technology's actual value proposition. But the load-bearing failure is that the reported results, taken at face value, do not demonstrate that Exp4 is any better than Exp2.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a co-design framework for a microgrid serving battery-electric truck charging, in which the capacities of battery, supercapacitor, flywheel, PV, and grid connection are optimized together with hourly dispatch over a 30-day synthetic period. Four storage configurations are compared in a Dutch distribution-center case study, and the authors report that the fully hybrid configuration achieves the lowest total cost, 1.96% below the battery-only baseline, at a higher capital cost. The framework is formulated as a linearized optimization problem that the authors call a MILP, with parameters from public sources and charging loads from previous work.","tokens_in":9085,"tokens_out":12666,"duration_ms":128086,"significance":"The topic is relevant and timely, and the paper has some strengths: the model couples sizing and operation, parameter values are taken from cited external reports, and the optimization pipeline (YALMIP/Gurobi) is standard. The synthetic-scenario construction via k-means and a Markov chain is clearly described. However, the numerical results as reported do not support the central claim: the best-configuration ranking is within the solver optimality gap, the tables and prose contradict each other, and the 30-day/hourly modeling choices cannot capture the sub-hourly transients that motivate supercapacitors and flywheels. Until these issues are resolved, the claimed cost saving is not established.","major_comments":[{"comment":"The claim that Experiment 4 achieves the lowest total cost is not supported by the reported data. Table 2 lists total costs of 22.387 for Exp2 and 22.386 for Exp4, a difference of 0.001 units, while the text reports a solver gap below 1%; a 1% gap on a cost of about 22.4 units is about 0.224 units, so the optimizer cannot resolve a 0.001-unit difference. The 1.96% saving relative to Exp1 is of the same order as the gap and therefore also not established.","section":"Table 2 and Abstract/§3"},{"comment":"The reported results are internally inconsistent. For Exp4, the text gives CapEx 2.520 and OpEx 20.287, whereas Table 2 gives CapEx 2.629 and OpEx 19.757. Moreover, using Table 2's own entries, CapEx + OpEx - EOL = 2.629 + 19.757 - 0.422 = 21.964, not the tabulated total cost 22.386; the text's numbers (2.520 + 20.287 - 0.422 = 22.385) come closer. The abstract's 2.6% higher initial investment also matches Table 2, not the text's 1.64%. These contradictions mean the supporting numbers are not reproducible.","section":"Section 3 vs Table 2"},{"comment":"The optimization is performed over a 30-day synthetic period, yet the objective is presented as a 20-year net present value. Equation (31) sums discounted yearly costs, but the paper never specifies how the 30-day simulation cost is annualized (e.g., by a factor of 365/30) or validated against annual totals, seasonal variability, or peak events. The extrapolation from one synthetic month to 20 years is therefore undescribed, and the resulting 1.96% saving cannot be interpreted as a long-run cost reduction.","section":"Section 2.6, Eq. (31), Table 1"},{"comment":"With τ=60 min, the model constrains only hourly average power and hourly energy transitions. The paper motivates supercapacitors and flywheels by their ability to handle rapid, sub-hourly fluctuations, but an hourly discretization cannot represent such transients; the C-rate constraints in (18) limit changes per hour, not instantaneous power. As a result, the cost differences among configurations with and without supercapacitors/flywheels are not anchored to the physical capability that these technologies are supposed to provide.","section":"Section 2.3, Table 1, Fig. 3"},{"comment":"The problem is called a mixed-integer linear program, but no integer or binary decision variables appear in the formulation; all variables in Problem 1 are stated to be real, and the k-means cluster assignments in Section 2.6 are pre-processing, not part of the optimization. The problem is therefore a linear program after the McCormick linearization, and the term 'MILP' is inaccurate.","section":"Section 2.5, Problem 1"}],"minor_comments":[{"comment":"'sold energy and purchased energy increasing slightly with Exp. 2' should refer to Exp. 3; Table 2 shows Exp3 sold and purchased energies higher than Exp2.","section":"Section 3, after Table 2"},{"comment":"'adding flywheels (Exp. 3 and Exp. 4) increases sold energy' is contradicted by Table 2, where Exp4 sold energy (234.53) is slightly below Exp1 (234.73).","section":"Conclusions"},{"comment":"The constraint has the quantifier '∀s ∈ S' but applies to the grid; it should be '∀k ∈ K' or explicitly to grid power.","section":"Eq. (10)"},{"comment":"The formula places J_cap_s inside the same parenthesis as the storage resale term; please clarify the intended expression and check the dimensions of the cycle-based resale value.","section":"Eq. (36)"},{"comment":"The table contains apparent typos: the second 'CPe' entry (0.001) is likely a different parameter, and 'Cop' is undefined; units are inconsistently written as 'k e' and 'ke'.","section":"Table 1"},{"comment":"There are numerous spelling errors (e.g., 'repsented', 'agreggation', 'Mantainance', 'pondered', 'disount', 'callibrated', 'exsting', 'comparsion', 'vehcles'); the paper needs a careful proofread.","section":"Throughout"}],"recommendation":"reject","confidential_remarks":"The central failure is not a disagreement with consensus but an internal numerical inconsistency: the reported optimality gap makes the Exp2/Exp4 ranking meaningless, and the table/prose accounting for Exp4 does not reconcile. A revision would need to re-solve with a much tighter gap or prove exactness, correct the table and prose, and redo the case study at a temporal resolution appropriate for fast storage. These are substantial changes to the evidence, so I do not see the current manuscript as salvageable by minor edits."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Useful co-design framework for hybrid storage in truck-charging microgrids, but the headline claim that the fully hybrid solution beats the others is not supported by the data you report. In Table 2, Exp2 and Exp4 have total costs 22.387 and 22.386 k€—a one-euro difference on a 22 k€ problem while the solver gap is 1% (roughly ±220 k€). That is noise, not a result. The abstract says 'best outcomes' but the optimizer cannot resolve those two configurations.\n\nThe genuinely new piece is the joint optimization of battery, supercapacitor, and flywheel sizing with operational dispatch for BET charging, using realistic schedules and market prices. The LP formulation with the lossless relaxation argument is clean, and the synthetic scenario generation via k-means and a Markov chain is standard. The authors use real data from PVGIS and ENTSO-E, and they are upfront about sensitivity to price assumptions. That is real work.\n\nSoft spots beyond the headline: the text contradicts Table 2 for Exp4 (CapEx 2.520 vs 2.629 k€, OpEx 20.287 vs 19.757 k€). That has to be fixed. More fundamentally, the hourly time step cannot represent the sub-hourly transients that justify supercapacitors and flywheels, so the 1.96% saving over battery-only is not anchored to the technology's actual value proposition. Also, the 30-day synthetic month is extrapolated to 20 years without any description of how that extrapolation works or validation that it preserves the relevant dynamics. Finally, labeling Problem 1 an MILP is wrong—all decision variables are real, and you rely on a lossless relaxation; call it an LP.\n\nNone of these are fatal to the framework. The co-design model is coherent and the case study is relevant. But the central claim as written is not substantiated. A serious referee should see this because the topic matters and the issues are correctable: fix the reporting, run a sensitivity analysis on solver tolerance, and either move to a finer time step or justify why hourly is adequate for power-dense storage.","headline":"Useful co-design framework for hybrid storage in truck-charging microgrids, but the headline result that the fully hybrid solution is best is not supported by the reported data—Exp2 and Exp4 are tied within solver tolerance and the text contradicts Table 2.","tokens_in":9524,"tokens_out":2551,"would_cite":false,"duration_ms":24435,"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":"For a truck-charging microgrid, hybrid battery-plus-fast storage cuts 20-year cost by 1.96% versus battery-only.","keywords":["hybrid energy storage","co-design","mixed-integer linear programming","microgrid","battery electric truck charging","supercapacitor","flywheel","net present value"],"falsifier":"Re-run the identical objective on the full three years of historical hourly data instead of the synthetic 30-day month: if the optimal storage mix changes or the hybrid advantage over battery-only disappears, the central claim fails. A sharper test is to solve one representative week at one-minute resolution; if the selected supercapacitor and flywheel capacities change substantially at that resolution, the hourly co-design was missing the very transients that justify them.","tokens_in":8481,"feed_emoji":"⚡","tokens_out":8141,"duration_ms":81194,"temperature":0.7,"pith_summary":"The paper argues that stationary storage for battery-electric truck charging should be designed and operated in a single optimization, because sizing and daily dispatch decisions are coupled. For a distribution-center microgrid in the Netherlands, the authors claim that a battery-only system is already competitive, but that adding supercapacitors and a flywheel lowers the 20-year total cost by 1.96% at a 2.6% higher initial investment. The result comes from solving a mixed-integer linear program with global optimality guarantees, using a synthetic representative month built from three years of solar, price, and charging data. A sympathetic reader would take the paper's core message to be that hybrid storage can pay for itself through lower operating cost and reduced grid dependence, without requiring a heuristic or sequential design process.","feed_headline":"Hybrid storage beats battery-only for truck charging by 1.96%","feed_subtitle":"Co-designing battery, supercapacitor, and flywheel storage with their operating strategy cuts cost and grid dependence.","key_machinery":"The central object is a mixed-integer linear program in which installed capacities of the battery, supercapacitor, flywheel, solar panels, and grid connection are optimized together with every hourly power dispatch decision. The formulation stays linear by introducing an auxiliary variable $q_{e,k}=E^{\\max}_e R_{e,k}$ for storage throughput, so the C-rate limit $R_{e,k} \\le R^M_e$ can be enforced without bilinear products; the paper states that this convex relaxation is lossless for selling factors in $(0,1]$. The optimization horizon is a synthetic 30-day month assembled by k-means clustering of historical days and a Markov chain over cluster transitions, with costs annualized over 20 years at a discount rate.","core_discovery":"On its own terms, the paper's central discovery is that the fully hybrid system—battery, supercapacitor, and flywheel together—is the optimal solution of a jointly optimized design-and-control problem for the studied truck-charging microgrid. It achieves the lowest total cost of ownership, about 1.96% lower than battery-only, and the lowest operating cost, while keeping energy sales to the grid nearly unchanged and slightly reducing purchased energy. The authors also show that battery-only has the lowest capital cost, making the choice a trade-off between upfront investment and long-term operating expenses. They frame this as evidence that co-design, rather than sequential sizing followed by scheduling, can reveal storage combinations that a battery-only design would miss.","pith_inferences":["The 1.96% saving is about the same size as typical uncertainty in storage capital costs, so under different price assumptions the ranking between hybrid and battery-only could reverse; the paper itself flags sensitivity to price assumptions.","An hourly time step may undervalue supercapacitors and flywheels, whose main benefit is sub-hourly response; a one-minute-resolution study of the same site could either strengthen the hybrid case or show that hourly operation is what actually drives cost.","The model sells energy back to the grid but does not price ancillary services such as frequency regulation; adding those revenue streams would likely improve the economics of fast storage further.","Because the charging profiles come from schedules optimized to limit peak consumption, the case study may be unusually friendly to battery-only storage; uncoordinated charging elsewhere could increase the measured benefit of fast storage."],"forward_implications":["Sites that can accept a 2.6% higher capital outlay can expect lower 20-year total cost and lower operating expenses from the fully hybrid storage system.","Adding supercapacitors alone or a flywheel alone each reduce total cost relative to battery-only, so partial hybridization is also a rational intermediate choice.","Because the MILP is solved with global optimality guarantees, the cost ranking among storage configurations is not an artifact of heuristic sizing or scheduling.","The framework transfers to other sites by replacing solar, price, and charging-load inputs, making the choice of storage mix a data-driven calculation rather than a rule of thumb."],"supporting_citations":[{"why":"Supplies the truck charging load profiles that define the demand the microgrid must serve.","marker":"Bertucci et al., 2024"},{"why":"Provides the historical solar generation data used to build the synthetic month.","marker":"ECJRC, 2025"},{"why":"Provides the electricity price time series used in the objective function.","marker":"ENTSO-E, 2025"},{"why":"Source of energy storage cost and performance parameters used in the case study.","marker":"Minear et al., 2018"},{"why":"Source of storage technology cost characterizations used in the case study.","marker":"Mongird et al., 2019"},{"why":"Source for stationary storage technology characteristics feeding the cost and efficiency inputs.","marker":"Kebede et al., 2022"},{"why":"Source for renewable power cost inputs, including PV capital cost.","marker":"IRENA, 2024"}],"fun_headline_variants":["Battery, supercaps, and flywheel together cut truck-charging cost","Co-design reveals hybrid storage beats battery-only for truck charging","Hybrid storage shaves 1.96% off truck-charging cost, at higher upfront cost","Optimal storage for truck charging: it's not just batteries","Supercapacitors and flywheels justify hybrid storage for truck charging"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The numbers depend on a synthetic 30-day month, generated by clustering three years of data and a Markov chain, being representative of 20 years of operation at an hourly resolution, including the fast power transients that supercapacitors and flywheels exist to handle; if that month misrepresents those transients, the chosen sizes and the 1.96% saving are unreliable.","fun_headline_variants_meta":{"raw":{"variants":["Battery, supercaps, and flywheel together cut truck-charging cost","Co-design reveals hybrid storage beats battery-only for truck charging","Hybrid storage shaves 1.96% off truck-charging cost, at higher upfront cost","Optimal storage for truck charging: it's not just batteries","Supercapacitors and flywheels justify hybrid storage for truck charging"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000542,"raw_usage":{"total_tokens":2576,"prompt_tokens":906,"completion_tokens":1670,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":522,"completion_tokens_details":{"reasoning_tokens":1570}},"tokens_in":522,"tokens_out":1670,"duration_ms":14152,"temperature":1.0,"reasoning_tokens":1570,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:43:51.448616+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the identical objective on the full three years of historical hourly data instead of the synthetic 30-day month: if the optimal storage mix changes or the hybrid advantage over battery-only disappears, the central claim fails. A sharper test is to solve one representative week at one-minute resolution; if the selected supercapacitor and flywheel capacities change substantially at that resolution, the hourly co-design was missing the very transients that justify them.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the truck charging load profiles that define the demand the microgrid must serve."},{"cited_title":"Photovoltaic geographical information system (pvgis)","cited_arxiv_id":null,"evidence_quote":"Provides the historical solar generation data used to build the synthetic month."},{"cited_title":"Transparency platform","cited_arxiv_id":null,"evidence_quote":"Provides the electricity price time series used in the objective function."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Source of energy storage cost and performance parameters used in the case study."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Source of storage technology cost characterizations used in the case study."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Source for stationary storage technology characteristics feeding the cost and efficiency inputs."},{"cited_title":"Renewable power generation costs in 2023","cited_arxiv_id":null,"evidence_quote":"Source for renewable power cost inputs, including PV capital cost."}],"review_version":1}