{"id":"42e867aa-a84e-40e5-b987-53ec8db8ae3b","arxiv_id":"1908.06352","paper_version":7,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Using a sequential two-layer model, the paper finds that smart HVAC demand-side management cuts a 5-node microgrid's annual investment and operation cost by 10.67 percent.","lead":"This paper links smart HVAC control with microgrid planning in a two-step model and applies it to a 5-node test network. It reports that smart cooling reduces total annual microgrid cost by about 10.7 percent compared with conventional setback control.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reported investment costs in Table 3 cannot be reconciled with Tables 2 and 4, so the headline 10.67% annual cost saving is not supported by the paper's own data.","rationale":"The paper's central quantitative claim is the 10.67% annual-cost reduction, which is computed from the Total Annual Cost entries in Table 3. The reader's conditional verdict focuses on the unvalidated cooling-load regression (Eq. 3). That is a real concern about the DSM layer's input accuracy. However, a more basic problem sits one step downstream: the Investment Cost column of Table 3 cannot be reproduced from the paper's own technology-cost table (Table 2) and optimal-capacity table (Table 4). A straightforward capital-cost calculation for Scenario I yields $14.28M, more than 180 times the reported $79,420 investment-cost entry; annualizing this over 25 years at zero interest still gives about $571,000 per year, about seven times the reported value. No capital-recovery factor or planning horizon M is given in Eq. 4, and the paper never states whether the Investment Cost column is annualized, one-time, or something else. If it is one-time, adding it to annual operation cost to form 'total annual cost' is dimensionally invalid. Under either reading, the 10.67% figure is not a well-defined, reproducible quantity from the submitted manuscript. This is an internal inconsistency, not a disagreement with community cost assumptions. The appropriate verdict is therefore UNVERDICTED: the authors need to document the cost metric and reconcile Tables 2 through 4 before the central claim can be evaluated. The reader's regression concern remains relevant and should also be addressed, but it is secondary to the cost-accounting inconsistency.","tokens_in":8283,"tokens_out":12308,"duration_ms":124731,"concrete_test":"Recompute the Investment Cost column of Table 3 from Tables 2 and 4 using the objective in Eq. 4 with an explicit capital-recovery formula. For example, take Scenario I capacities and costs: PV 1030×3000, ESS 320×600, CHP-MT 2000×3500, CHP-FC 1000×4000, then annualize over the stated lifetimes with a stated interest rate. If the resulting annualized investment cost is not within rounding of $79,420, and the Scenario I/II difference is not within rounding of $10,959, the 10.67% claim is not supported by the paper's own data. Also request the explicit definition of C_invd and C_invc in Eq. 4, including whether they are annualized, one-time, or net-present-value terms.","verdict_should_be":"UNVERDICTED","load_bearing_attack":"The central 10.67% saving is the difference between the two Total Annual Cost entries in Table 3, but the Investment Cost column cannot be derived from the paper's own cost and sizing data. Scenario I capacities in Table 4 are PV 1030 kW, CHP-MT 2000 kW, CHP-FC 1000 kW, ESS 320 kW. Applying Table 2's costs at face value gives 2000×3500 + 1000×4000 + 1030×3000 + 320×600 = $14,282,000 of capital. Even straight-line annualization over 25 years at 0% interest gives about $571,000 per year, not the reported $79,420. Scenario II, with PV 850 kW and ESS 240 kW, gives about $548,000 per year, for a difference of about $23,500 rather than the reported $10,959. Eq. 4 states C_invd + C_invc plus annual operating sums, but no capital-recovery factor, interest rate, or planning horizon M is specified. If 'investment cost' is instead a one-time cost, then adding it to annual operation cost to form 'total annual cost' is dimensionally invalid. Under either reading, the headline cost comparison in Table 3 is not reproducible from the manuscript's stated inputs. The reader's concern about unvalidated regression Eq. 3 is legitimate, but this cost-accounting inconsistency is more fundamental: even with perfect load profiles, the headline number is not supported by the paper's own cost and capacity tables.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a two-layer optimization model for microgrid planning with demand-side management. The first layer optimizes building zone temperature setpoints to minimize cooling electricity cost plus an occupant productivity penalty; the second layer solves a mixed-integer linear program for the microgrid's DER portfolio, sizing, placement, and operation, including investment and operation costs. The model is demonstrated on a 5-node, 12 kV microgrid with two medium-office and two mid-rise apartment buildings, comparing conventional setback control (Scenario I) with smart HVAC control (Scenario II). The authors report that smart control reduces the total annual cost from $242,711 to $216,807 (a 10.67% saving) and reduces installed PV and ESS capacities.","tokens_in":8615,"tokens_out":6548,"duration_ms":64522,"significance":"The integration of demand-side management with DER portfolio optimization is a relevant and under-addressed problem, and the paper's two-layer structure is a sensible way to expose the interaction between building-level control and microgrid investment decisions. If the reported 10.67% cost saving were established, it would be a useful demonstration that setpoint control can change optimal microgrid portfolios. The work builds on an existing RU-LESS optimization platform and applies established linearization methods for distribution networks and thermal networks, which are appropriate tools for this type of planning study. However, the manuscript as submitted does not provide enough information to verify either the demand-side model or the cost accounting, so the quantitative claims cannot currently be accepted as evidence.","major_comments":[{"comment":"The investment-cost entries in Table 3 cannot be reproduced from the capacities in Table 4 and the unit costs in Table 2. For Scenario I, the CHP-MT (2,000 kW x $3,500/kW = $7.0M) and CHP-FC (1,000 kW x $4,000/kW = $4.0M) alone exceed the reported investment cost by two orders of magnitude. Even if the total capital is amortized over 25 years at 0% interest, the annualized investment is approximately $571,000, or about $660,000 if the continuous-technology costs are also treated as one-time capital, not the reported $79,420. Eq. (4) adds C_invd and C_invc to annual operating sums, but no capital-recovery factor, interest rate, or planning horizon M is specified. If C_invd and C_invc are one-time costs, then the \"Total Annual Cost\" in Table 3 is dimensionally invalid; if they are annualized, the missing financial parameters prevent verification. The headline 10.67% saving and the 13.79% investment-cost reduction are therefore unsupported by the stated data. The authors should provide the complete cost-accounting model and recalculate the results.","section":"III-B, Eq. (4), Tables 2-4"},{"comment":"The cooling-electricity regression ASE_z^t = beta_0 + ... + epsilon is taken from reference [28] and applied to the four case-study buildings without re-estimation or validation. The paper reports no R-squared, no residual analysis, no out-of-sample test, and no sensitivity of the final 10.67% saving to the regression coefficients. Because the \"smart\" load profiles in Table 1, Scenario II are generated entirely by this model, any error in Eq. (3) propagates directly into the microgrid sizing and cost comparison. The authors should either re-estimate the regression on the actual case-study buildings, provide validation statistics, and report a sensitivity analysis around the regression coefficients, or clearly state the external validity limitations and qualify the quantitative conclusions accordingly.","section":"III-A, Eq. (3)"},{"comment":"The parameters of the DSM objective are not specified: the weights w_z,1 and w_z,2 in Eq. (1) and the occupant profit constant EP are never assigned numerical values, and the constraints of the first-layer optimization (comfort temperature bounds, setpoint ramp limits, and the optimization horizon) are not stated. Consequently, the optimal setpoint trajectories and the resulting load profiles are not reproducible from the manuscript. The authors should report these parameter values and the full constraint set for the DSM layer.","section":"III-A, Eq. (1)"}],"minor_comments":[{"comment":"The caption of Table 2, \"ANNUAL ELECTRICAL AND COOLING DEMAND,\" is identical to Table 1's caption and does not describe the table's content; it should read something like \"Technology cost and efficiency parameters.\"","section":"Table 2"},{"comment":"The summation and indexing notation in Eqs. (1) and (3) is garbled (e.g., \"6 t-1 N z-1\"); the sums over time and zones should be written with clear indices such as t=1..T and z=1..N.","section":"Eqs. (1) and (3)"},{"comment":"The header \"ALL NUMBERS ARE IN KW\" should be \"All numbers are in kW.\" Also, \"Displacement\" in the table title and in the Section III-B heading should be \"Placement.\"","section":"Table 4"},{"comment":"For the continuous technologies, the meaning of \"Fixed Cost ($/kW)\" and \"Variable Cost ($/kW)\" is unclear; the authors should state whether these are capital costs, annual fixed O&M costs, or energy-related costs, and specify their units.","section":"Table 2"},{"comment":"The Node 1 annual cooling usage of 12,958 MWh_th is more than eight times the Node 2 value of 1,497 MWh_th and appears implausible for a medium office building; the authors should check the value and the units.","section":"Table 1"}],"recommendation":"major_revision","confidential_remarks":"The cost-accounting inconsistency in Tables 2-4 is the most serious issue: it undermines the headline quantitative claim in its current form. The issue is, in principle, fixable by providing the missing economic parameters and recalculating, but the revision needs to be substantial rather than cosmetic. The paper also needs fuller specification of the DSM model before it can be considered a complete archival contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline result here — a 10.67% annual cost reduction from smart HVAC control on a 5-node microgrid — does not survive contact with the paper's own cost tables. Table 3 reports an investment cost of $79,420 for Scenario I. But applying Table 2's capital costs to Table 4's capacities gives roughly $14.28M of installed equipment (2000 kW × $3500 for the CHP-MT, 1000 kW × $4000 for the CHP-FC, 1030 kW × $3000 for PV, 320 kW × $600 for ESS). Even straight-line annualized over 25 years at zero interest, that's about $571k/year, seven times the reported figure. If the $79,420 is instead a one-time cost, then adding it to an annual operating cost is dimensionally invalid. Either way, the 10.67% saving cannot be reproduced from the stated data. That is a load-bearing problem: the entire quantitative conclusion sits on this table.\n\nWhat the paper does well is set up a sensible two-layer workflow: run a building-level cooling controller first, then feed the resulting load profiles into a microgrid portfolio optimization. The direction of the result is plausible — less cooling load should reduce required PV and ESS capacity — and the authors are open about building on their own RU-LESS platform and on Mashayekh et al.'s MILP formulation. That integration is the paper's actual contribution, and it's a legitimate one, if modest.\n\nThe weaker secondary issues the reader flagged are real too. Equation 3 is a regression model from prior work [28] with no coefficients, no fit statistics, and no validation on these buildings. The objective weights w_z,1, w_z,2 and the occupant profit EP in Eq. 1 are unspecified. No code or full input data are provided. So even the demand-side load profiles are not independently checkable.\n\nThe cost-accounting gap is the more fundamental problem, though. Fixing it might change the magnitude or even the sign of the claimed savings. The paper would need a clear statement of the capital-recovery factor, interest rate, planning horizon, and whether the investment cost is annualized or one-time. With those, a dedicated reader might reconcile the tables and the headline might hold; without them, the 10.67% is not a result.\n\nWho is this for? A microgrid planner could take away a qualitative lesson — DSM shifts optimal DER portfolios — but the quantitative claims should not be cited until the accounting is corrected. I would not cite it in its current form. It deserves a serious referee only if the authors are willing to do a major revision and supply the missing parameters and ideally the data. Send it back for that.","headline":"A plausible two-layer DSM-to-microgrid workflow, but the headline 10.67% saving does not survive the paper's own cost tables; the investment cost is off by an order of magnitude.","tokens_in":9170,"tokens_out":3090,"would_cite":false,"duration_ms":30127,"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":"Smart HVAC setpoint control cuts a five-node microgrid's total annual cost by 10.67 percent and lowers optimal PV and storage capacity.","keywords":["microgrid optimization","demand-side management","smart HVAC control","distributed energy resources","portfolio sizing and placement","mixed-integer linear programming","cooling load regression","techno-economic analysis"],"falsifier":"Re-run the two-layer optimization using cooling loads for the four buildings obtained from actual meter data or a validated building simulation while holding all other settings fixed; if the smart-control versus setback-control gap in total annual cost is not close to $25,904, or changes sign, the central claim is falsified. A more direct check is to compare the Eq. (3) regression predictions with measured cooling electricity use in the two office buildings and two apartment buildings over a cooling season and see whether the prediction error is smaller than the claimed saving.","tokens_in":8102,"feed_emoji":"⚡","tokens_out":7690,"duration_ms":74260,"temperature":0.7,"pith_summary":"This paper tries to establish that demand-side management is not a separate add-on to microgrid planning: the way buildings schedule their cooling setpoints changes the optimal portfolio, size, and placement of generation and storage. The authors build a two-layer model in which a smart HVAC controller first minimizes cooling electricity cost and comfort penalties, and a mixed-integer linear program then minimizes total investment plus operation cost of a five-node microgrid using the resulting load profiles. Comparing smart setpoint control with conventional setback control, they report that total annual cost falls from $242,711 to $216,807, a 10.67 percent saving, with lower installed PV and storage capacity. The significance, if true, is that software-side control of building loads can substitute for some capital investment in renewables and batteries.","feed_headline":"Smart HVAC setpoints cut microgrid annual cost 10.67%","feed_subtitle":"Linking cooling control to portfolio sizing shrinks PV and battery needs while saving $25,904 a year.","key_machinery":"The load-bearing mechanism is the two-layer coupling between a zone-level temperature setpoint controller and a microgrid investment model. Layer 1 chooses each zone's setpoint $T_z(t)$ to minimize weighted cooling electricity cost plus a human-productivity penalty, subject to comfort bounds and temperature ramp limits; cooling electricity consumption is represented by the multivariate linear regression in Eq. (3), which predicts air-conditioning electricity from zone energy rate, ambient temperature, lagged zone temperatures, time-of-day category, and neighboring zone temperatures. Layer 2 feeds the resulting hourly load profile into a mixed-integer linear program that minimizes investment plus operation cost, selecting discrete CHP units and continuous PV and ESS capacities under linearized power flow, voltage, cable current, and thermal balance constraints. The claimed 10.67 percent saving comes from running that same optimization on setback-control versus smart-control load profiles.","core_discovery":"The paper's central claim is that feeding demand-side management results into microgrid portfolio optimization produces a materially cheaper design than optimizing for loads produced by conventional HVAC setback control. In the demonstration on a 12 kV five-node network with two medium office buildings and two midrise apartments, smart cooling control lowers total annual microgrid cost from $242,711 to $216,807, reduces investment cost by 13.79 percent and annual operation cost by 9.15 percent, and shrinks aggregate PV capacity from 1030 kW to 850 kW and ESS capacity from 320 kW to 240 kW while CHP capacity is unchanged. The authors present this as evidence that the optimal microgrid portfolio depends on how buildings are controlled, so DSM should be an input to, not an afterthought of, microgrid sizing and placement.","pith_inferences":["The quantitative saving is probably climate- and rate-specific; repeating the two-layer model in a heating-dominated climate or with different demand charges would likely change the size of the saving while preserving its direction for cooling loads.","A natural testable extension is to run the same optimization on metered load data from real buildings; if the cooling regression is re-estimated on those buildings, the reported cost gap may shrink or grow.","Because smart control shifts load timing as well as total energy, the same method could be used to evaluate whether DSM can defer feeder or transformer upgrades in addition to reducing PV and battery capacity."],"forward_implications":["If the result holds, the same microgrid constraints are met with 180 kW less PV and 80 kW less storage, so demand-side management acts as a capital-cost-reducing resource rather than only an operating lever.","Because CHP capacities are identical in both scenarios, the cost saving appears to come from shaving electric cooling loads that would otherwise drive PV and battery investment, not from resizing thermal generation.","The 10.67 percent total annual cost reduction, $25,904 per year, gives a concrete dollar value for smart HVAC control in a cooling-dominated five-node setting.","Peak cooling demand falls at every node in the case study, so grid imports and peak-hour network stress drop under smart setpoint control."],"supporting_citations":[{"why":"Supplies the data-driven cooling-electricity regression (Eq. 3) used to generate the smart-control load profiles.","marker":"[28]"},{"why":"Provides the linear power-flow approximation used for bus voltage and net injected power constraints in the microgrid layer.","marker":"[30]"},{"why":"Provides the MILP formulation and linearized cable-current and voltage constraints on which the microgrid portfolio optimization is built.","marker":"[31]"},{"why":"Supplies the original bus-voltage linearization that the microgrid model enhances with cable-current constraints.","marker":"[32]"},{"why":"Supplies the linear approximation of thermal losses used in the heating and cooling balance constraints.","marker":"[33]"},{"why":"Supplies the human-performance-versus-temperature curve used in the DSM objective's comfort penalty.","marker":"[29]"}],"fun_headline_variants":["Smart HVAC setpoints cut microgrid costs 10.67%","DSM-integrated microgrid design saves $25,904 a year","Linking cooling control to microgrid sizing reduces PV, ESS","Optimal microgrid portfolio depends on thermostat control","Two-layer optimization cuts microgrid annual cost 10.67%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The case study's smart-control load profiles come from a cooling-electricity formula fitted in earlier work and applied to these four buildings without being re-estimated or checked against measured data; if that formula mispredicts cooling load for these buildings, the 10.67 percent annual cost saving does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Smart HVAC setpoints cut microgrid costs 10.67%","DSM-integrated microgrid design saves $25,904 a year","Linking cooling control to microgrid sizing reduces PV, ESS","Optimal microgrid portfolio depends on thermostat control","Two-layer optimization cuts microgrid annual cost 10.67%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000239,"raw_usage":{"total_tokens":1513,"prompt_tokens":939,"completion_tokens":574,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":555,"completion_tokens_details":{"reasoning_tokens":487}},"tokens_in":555,"tokens_out":574,"duration_ms":5528,"temperature":1.0,"reasoning_tokens":487,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:47:52.591674+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the two-layer optimization using cooling loads for the four buildings obtained from actual meter data or a validated building simulation while holding all other settings fixed; if the smart-control versus setback-control gap in total annual cost is not close to $25,904, or changes sign, the central claim is falsified. A more direct check is to compare the Eq. (3) regression predictions with measured cooling electricity use in the two office buildings and two apartment buildings over a cooling season and see whether the prediction error is smaller than the claimed saving.","supporting_citations":[{"cited_title":"Distributed air conditioning control in commercial buildings based on a physical -statistical approach,","cited_arxiv_id":null,"evidence_quote":"Supplies the data-driven cooling-electricity regression (Eq. 3) used to generate the smart-control load profiles."},{"cited_title":"On the existence and linear approximation of the power flow solution in power distribution networks,","cited_arxiv_id":null,"evidence_quote":"Provides the linear power-flow approximation used for bus voltage and net injected power constraints in the microgrid layer."},{"cited_title":"A mixed integer linear programming approach for optimal DER portfolio , sizing , and placement in multi -energy microgrids,","cited_arxiv_id":null,"evidence_quote":"Provides the MILP formulation and linearized cable-current and voltage constraints on which the microgrid portfolio optimization is built."},{"cited_title":"A mixed - integer LP model for the reconfiguration of radial electric distribution systems considering distributed generation,","cited_arxiv_id":null,"evidence_quote":"Supplies the original bus-voltage linearization that the microgrid model enhances with cable-current constraints."},{"cited_title":"Structural and operational optimisation of distributed energy systems,","cited_arxiv_id":null,"evidence_quote":"Supplies the linear approximation of thermal losses used in the heating and cooling balance constraints."},{"cited_title":"Effect of Temperature on Task Performance in Offfice Environment,","cited_arxiv_id":null,"evidence_quote":"Supplies the human-performance-versus-temperature curve used in the DSM objective's comfort penalty."}],"review_version":1}