{"id":"6ed5a515-e369-4577-8ebb-0ba49affe901","arxiv_id":"1908.11197","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A bi-level real-time-pricing model for isolated microgrids is shown to shift EV charging to off-peak hours, but the peak-shaving effect is a direct consequence of the assumed price formula and the model ignores EV time-window constraints.","lead":"This paper proposes a bi-level optimization model for scheduling an isolated microgrid with renewables and electric-vehicle charging, solved with a hybrid JAYA-interior-point algorithm. It reports that demand response of EVs shifts charging to off-peak hours and reduces costs for both the microgrid and EV users.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lower-level EV model omits arrival/departure windows, so the reported peak shaving may be an artifact of charging EVs at times when they are not present.","rationale":"The lower-level EV charging model as written in Eqs. (20)-(23) is a linear program that minimizes charging cost subject only to aggregate power capacity, per-EV power bounds, and a cumulative energy bound. Since the real-time price in Eq. (24) is monotonically increasing in total load, the optimizer's unconstrained freedom to place EV load at the single global demand minimum—or to distribute it arbitrarily across all 24 hours regardless of when the EV is plugged in—makes the valley-filling outcome of Fig. 10 nearly inevitable. The paper itself reports arrival/departure times and claims EVs leave when charged (Section II-B, Fig. 4), so the absence of any availability constraint is a genuine modeling error, not a simplification disclosed by the authors. This error is load-bearing: without it, the demonstration of active EV participation and peak shaving collapses. The reader's weakest_assumption identifies exactly this, and I agree. Secondary issues (heuristic algorithm without optimality proof, single scenario, reliance on Eq. (24)'s price-load proportionality) are real but would not alone change the verdict. The concrete test of adding arrival/departure window constraints would settle whether the claimed benefit survives a realistic model.","tokens_in":15090,"tokens_out":4903,"duration_ms":49956,"concrete_test":"Add to the lower-level model the constraints P_i,t^EV = 0 for t < t_i,arr and t > t_i,dep, where t_i,dep is taken from Fig. 4 (or, if departure is 'immediately once charged', t_i,dep = t_i,arr + T_i,CH from Eq. (6)). Re-run Case 2 with the same parameters and compare the resulting load curve to Fig. 10 and the strategy costs to Fig. 5. If the peak-to-valley reduction shrinks or the MG/EV cost trade-off shifts materially, the reported demand-response benefit is an artifact of unconstrained charging times.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—that EV demand response achieves peak shaving and balances MG/EV interests—rests on the lower-level EV charging model in Eqs. (20)-(23). The paper generates arrival and departure times in Section II-B and Fig. 4, and states that EVs leave once charged to expected capacity, but the lower-level constraints do not enforce that an EV's charging power P_i,t^EV is zero outside its arrival-to-departure window. Eq. (21) only caps aggregate charging by available MG power; Eq. (22) only imposes per-period min/max power limits; Eq. (23) only bounds cumulative energy. Nothing in (20)-(23) prevents the optimizer from charging an EV at 03:00 when it arrives at 18:00, or from spreading a single EV's charge over all 24 hours. Because the real-time price in Eq. (24) is proportional to (P_EV,t + P_L,t), the lower-level LP naturally pushes EV load into the global off-peak hours; without availability constraints this is exactly the behavior shown in Fig. 10 and the basis for the 'peak shaving' conclusion. The modeling omission therefore directly manufactures the headline result, and the claimed 'win-win' cost balance in Fig. 5 is not established for realistic EV behavior.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a bi-level programming model for scheduling an isolated microgrid with renewable generation and electric-vehicle charging under real-time pricing. The upper level minimizes microgrid net operating cost subject to probabilistic spinning-reserve constraints; the lower level minimizes EV charging cost. A real-time price is computed by Eq. (24) as a linear function of total load plus EV load, and the bi-level problem is solved by an iterative JAYA-interior-point method. A 24-hour case study with 20 EVs compares three strategies and two demand-response cases, reporting that the joint strategy reduces both microgrid operating cost and EV charging cost and achieves peak shaving.","tokens_in":15345,"tokens_out":4509,"duration_ms":47005,"significance":"If the results were valid, the paper would offer a practical framework for coordinating EV demand response with renewable uncertainty in isolated microgrids, with a reusable heuristic solver. The paper deserves credit for modeling the probabilistic reserve constraint via sequence operation theory and for presenting a complete case study with multiple strategies and sensitivity cases. However, the central qualitative claim—that EV demand response achieves peak shaving and a win-win outcome—is not supported by the model as formulated, because the lower-level EV model lacks temporal availability constraints and the pricing rule cannot represent supply-demand dynamics. The reported results are therefore likely an artifact of the modeling assumptions, which substantially reduces the significance of the contribution.","major_comments":[{"comment":"The lower-level EV charging model omits the arrival and departure time windows that are generated in Section II-B and displayed in Fig. 4. Eq. (22) imposes only per-period min/max charging power limits, and Eq. (23) only bounds cumulative energy over the scheduling horizon; nothing prevents P_i,t^EV from being positive at hours when EV i is not connected to the charging station. Because the real-time price in Eq. (24) is monotonically tied to P_t^EV + P_t^L, the lower-level cost minimization will artificially shift all EV charging to the global off-peak hours. The peak-shaving result in Fig. 10 and the 'win-win' cost comparison in Fig. 5 therefore follow from unrealistic flexibility rather than from the proposed demand-response mechanism. This is a load-bearing modeling omission, not a minor detail: without availability constraints, the model permits charging an EV at 03:00 when it arrives at 18:00, so the central claim of the paper is not established for realistic EV behavior.","section":"Section III-B, Eqs. (20)-(23)"},{"comment":"The real-time price formula depends only on the sum of EV charging power and original load, normalized by a reference power, with no term representing renewable generation, system marginal cost, or the operating state of the microgrid. The paper states in Section III-C that the price reflects the dynamic relationship between supply and demand, but the supply side (PV, WT, MT, ESS) does not enter Eq. (24) at all. As a result, the proposed 'demand response' is essentially an incentive to shift load toward periods of low total load, and the claimed coordination between EV demand response and renewable generation uncertainty is not embodied in the pricing mechanism. The case-study conclusion that demand response helps consume renewable generation is therefore not supported by the model.","section":"Section III-C, Eq. (24)"},{"comment":"The JAYA-IPM solution method is not validated. The termination criterion in Step 12 is only whether the current iteration exceeds a pre-defined maximum number of iterations, and the parameter settings are ambiguous: Section V-A sets N_maxiter to 20 while also stating a JAYA maximum of 1500 iterations, without clarifying which limit applies to the bi-level loop. The paper does not compare JAYA-IPM with an exact or benchmark solver, does not report optimality gaps, and does not test convergence on the bi-level model. Since the quantitative comparisons among Strategies 1-3 in Fig. 5 and the demand-response cases in Figs. 8-10 depend on the quality of the optimized solutions, the reader cannot assess whether the reported costs and schedules are meaningful.","section":"Section IV, Steps 4-15, and Section V-A"}],"minor_comments":[{"comment":"The off-peak interval is listed as '00:60-07:00', which is not a valid time interval; it should presumably be '00:00-07:00'.","section":"Table 3"},{"comment":"The paper states that EVs leave the charging station immediately once charged to expected capacity, and Fig. 4 plots EV arrival and departure times; however, these departure times are never used in the lower-level optimization, which is confusing and should be clarified.","section":"Section II-B and Fig. 4"},{"comment":"Eq. (6) defines the charging time T_i,CH in terms of rated power and charging efficiency, but the lower-level model uses only an aggregate P_t^EV; the relationship between per-EV charging power P_i,t^EV and the aggregate variable is not stated, and Eq. (23) does not connect the cumulative energy bound to T_i,CH.","section":"Eq. (6) versus Eqs. (20)-(23)"},{"comment":"The manuscript contains numerous typographical and grammatical errors (for example, 'is it has become', 'considerate both'), which should be corrected in any revision.","section":"Throughout"},{"comment":"In Eq. (11), SOC_t is treated as stored energy with units of kWh, while SOC elsewhere denotes a dimensionless state of charge; the notation should be clarified to avoid unit inconsistency.","section":"Eq. (11)"}],"recommendation":"reject","confidential_remarks":"The paper is very closely related to the authors' earlier work [19], which uses the same bi-level structure, the same real-time price formula, and the same JAYA-IPM solution method for an EV battery-swapping station. The incremental contribution here is the replacement of battery swapping with direct EV charging, but the modeling of EV availability is less careful than in [19]. The editor may wish to check the novelty claim against [19] and consider whether the incremental contribution is sufficient for the journal's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Li and Li propose a bi-level scheduling model for an isolated microgrid with EVs responding to real-time prices, solved by a JAYA-interior point method. The paper is clearly written and the SOT-based deterministic transformation of the probabilistic reserve constraint is a clean way to avoid Monte Carlo; that piece is genuinely useful. The idea of extending their earlier battery-swapping-station work [19] to EV charging is a reasonable incremental step.\n\nThe problem is at the core. The lower-level EV model in Eqs. (21)-(23) never enforces the arrival/departure windows that Section II-B generates and Fig. 4 plots. Nothing stops the optimizer from charging an EV at 3 a.m. when it does not arrive until 18:00. That is a load-bearing omission. Compounding it, the real-time price in Eq. (24) is set proportional to the sum of EV load and original load. The lower-level cost minimization then naturally pushes charging into hours where the combined load is low, so the peak shaving shown in Fig. 10 is essentially manufactured by the pricing rule. The paper presents this as a demonstrated benefit, but it is a direct consequence of the model design.\n\nThe other soft spots are minor by comparison but still real. The JAYA-IPM hybrid is asserted to be sufficient and efficient without any benchmark against a standard solver, so the computational claims are unsupported. The case study is a single random scenario with no error bars, data, or code, which is thin evidence for the quantitative comparisons. The authors do acknowledge some limitations in the conclusion (no discharging, no waiting time), which is more than many papers do, but they miss the availability issue.\n\nNet: the framework is not novel enough to outweigh the missing constraint, and the headline 'win-win' result is not established for realistic EV behavior. I would not send this to peer review as is. The fix is clear—add arrival/departure windows to the lower level, rework the pricing to depend on system marginal cost rather than just load, and benchmark the solver. If those things happen, the paper might become a solid incremental contribution. As it stands, it is a useful example of how pricing rules and missing constraints can pre-determine the outcome.","headline":"The missing availability constraints in the lower-level EV model make the peak-shaving result an artifact, so the paper's central claim is not established.","tokens_in":15873,"tokens_out":3616,"would_cite":false,"duration_ms":38163,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that a two-level scheduling model with real-time pricing can use electric-vehicle demand response to shave peaks in isolated microgrids while cutting costs for both the microgrid operator and EV users.","keywords":["microgrid scheduling","demand response","electric vehicles","renewable generations","bi-level programming","real-time pricing","JAYA algorithm","sequence operation theory"],"falsifier":"Run the same case study with an added constraint that EV i can only charge between its sampled arrival and departure times shown in the paper's Fig. 4, with each EV's total energy requirement fixed, and compare the load curves and costs; if the peak-to-valley improvement and the joint cost reductions shrink or vanish, the central result depends on the missing time-window constraint.","tokens_in":14819,"feed_emoji":"⚡","tokens_out":7216,"duration_ms":68216,"temperature":0.7,"pith_summary":"The paper tries to show that demand response of electric vehicles, driven by real-time prices, can be incorporated into the scheduling of an isolated microgrid with wind, solar, and storage, and that doing so benefits both sides. It formulates the problem as a bi-level program: the upper level minimizes the microgrid's net operating cost, the lower level minimizes EV users' charging cost, and a real-time price signal links the two levels. A hybrid solution algorithm combining the JAYA heuristic with an interior-point method is proposed. On a test microgrid, the joint strategy reduces the peak-to-valley load difference and improves both the MG net operating cost and the EV charging cost compared with strategies that optimize only one side. The result matters because isolated microgrids need economical ways to absorb renewable variability and EV load without access to a main grid.","feed_headline":"Bi-level pricing cuts EV charging cost and shaves microgrid peaks","feed_subtitle":"Simulated isolated microgrid shows real-time prices move EV load off-peak and lower both sides' costs","key_machinery":"The load-bearing structure is the bi-level program itself, coupled by a real-time price rule: after the lower level returns an EV charging plan, the upper level sets the price proportional to the ratio of the combined EV and original load to a reference load, and the plan is re-optimized until a joint criterion stops improving. Supporting machinery includes sequence operation theory, which turns probabilistic renewable outputs into discrete sequences and converts the chance constraint on spinning reserve into a deterministic 0–1 constraint, and the JAYA-interior-point hybrid algorithm, where JAYA searches the upper level and the interior-point method solves the lower level.","core_discovery":"The central claim is that a bi-level scheduling model with real-time pricing coordinates EV charging with microgrid operation so that EV users, pursuing only their own charging-cost minimization, end up shifting load away from peak periods and lowering the microgrid's net operating cost as well. In the reported test system, the joint optimization strategy outperforms the strategy that ignores EV costs on EV charging cost, and outperforms the strategy that ignores MG revenue on MG net operating cost; the load curves show reduced EV charging in peak-load periods and increased charging in off-peak periods. The authors frame this as a coordination of renewable generation uncertainty and EV demand response in a single scheduling model.","pith_inferences":["An extension the paper does not pursue: the real-time price rule is a purely load-proportional signal, so comparing it with a price that also reflects renewable surplus or storage state would reveal whether the peak-shaving result is driven by the price shape or by the model's assumed EV flexibility.","The reported results likely depend on the lower-level model allowing EVs to charge in any hour; constraining each EV to its arrival–departure window would give a stiffer test of the peak-shaving claim.","The same bi-level template could be applied to other price-responsive resources, such as water heaters or home batteries, where the lower level minimizes a user's energy bill and the upper level manages a distribution feeder.","The joint objective treats MG and EV interests symmetrically; a different weighting would shift the compromise, so the 'win–win' conclusion is tied to that particular squared-deviation choice."],"forward_implications":["In the reported test case, joint optimization lower the MG net operating cost compared with the strategy that ignores MG revenue, and lowers EV charging cost compared with the strategy that ignores EV costs.","The real-time price mechanism shifts EV charging out of peak-load periods, reducing the peak-to-valley load difference in the microgrid.","The chance-constraint transformation handles renewable uncertainty without Monte Carlo simulation, keeping the bi-level solution process tractable.","The proposed framework yields a single schedule that balances both parties' interests, making it usable as a negotiation tool between a microgrid operator and an EV charging-station operator."],"supporting_citations":[{"why":"Supplies the isolated-microgrid model, ESS and MT constraints, renewable probability models, and the chance-constraint transformation the upper level builds on.","marker":"[2]"},{"why":"Provides the real-time-pricing-based bi-level scheduling template, the joint optimization criterion, and test-system data used in the case study.","marker":"[19]"},{"why":"Provides the national travel survey from which EV arrival times and daily mileage distributions are taken.","marker":"[30]"},{"why":"Supplies the normal-distribution model for the initial EV state of charge at arrival.","marker":"[31]"},{"why":"Supplies sequence operation theory, whose addition-type convolution produces the joint renewable-output probability sequence and enables deterministic conversion of the chance constraint.","marker":"[34]"},{"why":"Supplies the JAYA algorithm update rule used to solve the upper-level MG scheduling model.","marker":"[35]"},{"why":"Supplies the interior-point method used to solve the lower-level EV charging model.","marker":"[37]"}],"fun_headline_variants":["Bi-level pricing trims EV costs and microgrid peak load","Real-time prices guide EVs off-peak in isolated microgrids","Two-level scheduling cuts EV charging and microgrid peaks","EV demand response shaves peaks via bi-level pricing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that EV charging power can be scheduled in any hour: the lower-level constraints never limit each EV to its sampled arrival-to-departure window, so the flexibility used to create the peak-shaving result is not tied to the actual times EVs are present.","fun_headline_variants_meta":{"raw":{"variants":["Bi-level pricing trims EV costs and microgrid peak load","Real-time prices guide EVs off-peak in isolated microgrids","Two-level scheduling cuts EV charging and microgrid peaks","EV demand response shaves peaks via bi-level pricing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000688,"raw_usage":{"total_tokens":3045,"prompt_tokens":799,"completion_tokens":2246,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":415,"completion_tokens_details":{"reasoning_tokens":2178}},"tokens_in":415,"tokens_out":2246,"duration_ms":17326,"temperature":1.0,"reasoning_tokens":2178,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:45:38.763494+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same case study with an added constraint that EV i can only charge between its sampled arrival and departure times shown in the paper's Fig. 4, with each EV's total energy requirement fixed, and compare the load curves and costs; if the peak-to-valley improvement and the joint cost reductions shrink or vanish, the central result depends on the missing time-window constraint.","supporting_citations":[{"cited_title":"Optimal scheduling of an isolated microgrid with battery storage considering load and renewable generation uncertainties ,","cited_arxiv_id":null,"evidence_quote":"Supplies the isolated-microgrid model, ESS and MT constraints, renewable probability models, and the chance-constraint transformation the upper level builds on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the real-time-pricing-based bi-level scheduling template, the joint optimization criterion, and test-system data used in the case study."},{"cited_title":"2009 national household travel survey ,","cited_arxiv_id":null,"evidence_quote":"Provides the national travel survey from which EV arrival times and daily mileage distributions are taken."},{"cited_title":"Performance evaluation of an EDA -based large-scale plug -in hybrid electric vehicle charging algorithm ,","cited_arxiv_id":null,"evidence_quote":"Supplies the normal-distribution model for the initial EV state of charge at arrival."},{"cited_title":"Dependent discrete convolution based probabilistic load flow for the active distribution system,","cited_arxiv_id":null,"evidence_quote":"Supplies sequence operation theory, whose addition-type convolution produces the joint renewable-output probability sequence and enables deterministic conversion of the chance constraint."},{"cited_title":"Application of EOS -ELM with binary Jaya - based feature selection to real-time transient stability assessment using PMU data ,","cited_arxiv_id":null,"evidence_quote":"Supplies the JAYA algorithm update rule used to solve the upper-level MG scheduling model."},{"cited_title":"Optimal planning of electric - vehicle charging stations in distribution systems ,","cited_arxiv_id":null,"evidence_quote":"Supplies the interior-point method used to solve the lower-level EV charging model."}],"review_version":1}