{"id":"febe1be4-d602-4b0e-b983-428f802baeed","arxiv_id":"1908.00481","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Wider comfort bounds in an economic MPC smart-home model shift 6-10% of annual electricity use to low-price periods and cut annual cost by 16-34%, depending on heating system and insulation.","lead":"This paper simulates a smart home that uses a controller to plan heating and appliance use one day ahead based on electricity prices and weather. It finds that letting the temperature swing a few degrees shifts 6-10 percent of electricity use to cheap hours and cuts annual bills by up to a third.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central savings and load-shift claims rest on perfect day-ahead forecasts; no forecast-error robustness test is provided.","rationale":"The paper's compact state-space formulation (Section III) is a reasonable extension of prior building-MPC work, and the simulation results in Table III are internally consistent: annual cost decreases and the low-price share increases as comfort bounds widen. The qualitative finding that FH and HVAC systems respond differently to insulation (Table V) is a useful contribution. The central weakness is that the quantitative claims in Section V are produced under a perfect-foresight, non-robust MPC simulation. The reader's conditional verdict is appropriate: the paper should be accepted subject to the authors either adding a forecast-error sensitivity study or tempering the quantitative conclusions to reflect that they are ideal upper bounds. I do not see an internal inconsistency that would justify rejection. The proposed concrete test would settle whether the specific savings and shifts survive realistic forecast uncertainty. If the test passes, the claims are strengthened; if not, the paper would need to reframe its conclusions as potential rather than expected.","tokens_in":10777,"tokens_out":15861,"duration_ms":166732,"concrete_test":"Re-run the one-year simulation for the FH and HVAC extraflex/PI-CB cases with a receding-horizon MPC: at each 15-min step, solve the day-ahead problem using a noisy forecast of the remaining ambient temperature (e.g., Gaussian error, sigma = 1.5 C), occupancy, and hot-water demand (e.g., +-20% multiplicative noise) for the rest of the day, while the building evolves under the true historical disturbance profiles; take only the current step's control, then re-solve. Compare the resulting annual electricity cost and low-price consumption share with Table III. If the extraflex-vs-noflex cost saving falls by more than 30% or the low-price share increase drops below 5 percentage points, the headline claims are not robust to forecast error.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section V's quantitative conclusions (cost savings of 16–34%; low-price consumption share rising from 54.5–57.2% to 64.1–64.7%) are obtained with an economic MPC that is simulated using perfect day-ahead knowledge of ambient temperature, occupancy, hot-water demand, and prices. Section IV-A states 'we use a look-ahead window of one day' and the data are the historical profiles of [18]; the controller therefore never experiences forecast error, and the reported annual numbers are ideal-planning results, not closed-loop outcomes under uncertainty. The paper contains no sensitivity analysis that perturbs the forecasts. Under imperfect weather or demand forecasts, the MPC will preheat or coast at the wrong times, so the simulated cost savings and the shift of consumption into low-price periods are optimistic upper bounds. Since the paper's central claim is that widening comfort bounds delivers these specific savings in practice, this is the most load-bearing assumption; the unvalidated linear model (Eq. 4) is also a concern, but forecast error directly attacks the reported magnitudes.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a compact economic model predictive control (MPC) formulation for a single-zone smart household that includes a five-state thermal model (room, floor, floor-water, refrigerator, water heater), two alternative space-heating systems (water-based floor heater and HVAC), lighting, and interruptible appliances with discrete scheduling constraints. The authors simulate one year of operation under perfect day-ahead forecasts, comparing three flexibility levels (noflex, flex, extraflex) and two comfort-bound strategies (price-independent and price-dependent), and report annual electricity cost, temperature-bound violations, and the share of consumption occurring in low-price periods. They conclude that widening comfort bounds produces cost savings of 16–34% depending on heating system and strategy, and shifts 6–10 percentage points more of annual consumption into low-price periods. They also analyze the effect of the building's heat-transfer coefficient on these outcomes.","tokens_in":10905,"tokens_out":6286,"duration_ms":66712,"significance":"If the reported magnitudes hold under realistic operating conditions, the paper would provide useful estimates of the demand-side flexibility potential of smart homes, a topic of current interest for distribution system operators and aggregators. The paper's strengths are its integrated formulation that combines thermal dynamics, appliance scheduling, and comfort constraints in a single optimization model; the explicit modeling of uninterruptible loads with variable power cycles; the comparison of two space-heating technologies; and the parametric study of insulation level. The authors also provide a data link for reproducibility. However, the quantitative claims rest on a perfect-forecast assumption, an unvalidated linear thermal model, and a single year of data, which limits the external validity of the specific percentages; the contribution is best viewed as an idealized upper-bound analysis rather than a robust prediction of achievable savings.","major_comments":[{"comment":"The simulations use perfect day-ahead knowledge of ambient temperature, solar radiation, occupancy, hot-water demand, and electricity prices, as stated in Section IV ('we use a look-ahead window of one day' and the historical profiles of [18]). No sensitivity analysis with forecast error is provided. Since economic MPC under imperfect forecasts will preheat or shift load at suboptimal times, the central quantitative claims in Section V (cost savings of 16–34% and the increase in low-price-period consumption share from 54.5–57.2% to 64.1–64.7%) are idealized upper bounds. To support the practical framing, the authors should include robustness tests with, for example, additive noise on prices, temperature, and occupancy forecasts, or with a stochastic MPC formulation.","section":"Section IV and Section V"},{"comment":"The five-state linear thermal model in Eq. (4) is adopted from [8] and [18] without validation against measured data for the simulated household. The only structural sensitivity study, in Section IV-B, varies the room-to-ambient heat transfer coefficient UAr,a while all other parameters and the model structure (constant COP, no humidity, no thermal solar-gain term) remain fixed. Because the reported savings and load-shift percentages are specific to this assumed model, the paper should either validate the model against data or clearly state that the results are illustrative of the model, and ideally include sensitivity analysis over other thermal parameters.","section":"Section III, Eq. (4), and Section IV"},{"comment":"The price-responsiveness metric is defined by an arbitrary threshold: low-price periods are those where the annual normalized price is lower than 0.5. The central claim in Section V that the share of building consumption in low-price periods rises from 54.5–57.2% to 64.1–64.7% depends on this threshold. No sensitivity to the threshold value is reported, so it is unclear whether the observed shift is robust or partly an artifact of the chosen cutoff. The authors should report the consumption share for a range of thresholds or provide a price-elasticity curve.","section":"Section IV-A (low-price-period definition)"}],"minor_comments":[{"comment":"The statement that insulated households with HVAC systems 'may lead to cost savings up to 50% approximately' is not directly traceable to any table or figure; please specify the exact comparison and provide the supporting number.","section":"Section V"},{"comment":"The description 'We run daily simulations with 15-min time steps for one year' combined with 'we use a look-ahead window of one day' is ambiguous: it is unclear how the state is passed between consecutive daily optimizations and whether a full-year receding-horizon MPC is being approximated. Please clarify.","section":"Section IV"},{"comment":"The comfort light levels (100 and 10000 lux) are stated in the text but not listed in Table II, which otherwise summarizes the comfort constraints; adding them would make the table self-contained.","section":"Section IV and Table II"},{"comment":"The manuscript contains typographical artifacts such as 'increasin g' and 'add ition' in the Abstract; a careful proofread is needed.","section":"Abstract and throughout"}],"recommendation":"major_revision","confidential_remarks":"This is a competently executed simulation study with a clear formulation and reproducible data, but the central quantitative results are not yet supported for realistic deployment because the MPC operates under perfect forecasts and the thermal model is not validated. The absence of any uncertainty analysis is the key gap; typical reviewers in the smart-grid community will expect at least a forecast-error sensitivity study. Please also note that the auxiliary data are provided via a goo.gl short link, which is not acceptable for archival purposes; the data should be deposited in a permanent repository."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful thing here is the integrated model: five-state thermal dynamics for two heating systems plus refrigerator, water heater, lighting, and phase-dependent uninterruptible loads in one economic MPC, run over a full year. The discrete phase scheduling for appliances is a real addition to the building-MPC literature. The sensitivity analysis of comfort bands, price-dependent bounds, and insulation is also done carefully. On those terms the paper is a solid extension of Halvgaard and related work. Cost savings of 16–34% and a 6–10 point shift into low-price periods are plausible within the model.\n\nThe soft spots are real but not disqualifying. The building model is a linear grey-box taken from earlier papers, not validated against measured data. The MPC is run with perfect day-ahead forecasts of weather, occupancy, hot-water demand, and prices. That matters: the reported numbers are ideal-planning results, not closed-loop outcomes under uncertainty. A forecast-error test would tell us how much of the 16–34% survives. The low-price threshold is arbitrary, and the results come from one household, one weather year, and one price series, so the quantitative claims should be read as case-study results, not general bounds. I also note the main direction is built into the objective: widening comfort bounds has to lower cost in an economic MPC. What is not built in is the magnitude, and that is what the paper estimates.\n\nThe paper does not oversell much; the conclusions mostly stay close to the simulations. The auxiliary data link is good practice. The price-dependent comfort bounds result, where flexibility costs more but reduces discomfort, is a nice nuance.\n\nWho is this for: people building residential demand-response models or needing a compact benchmark formulation. I would not treat the quantitative savings as empirical evidence. I would treat the model and sensitivity setup as a usable starting point.\n\nRecommendation: serious referee, yes. The main revision request should be a forecast-error sensitivity study and an explicit caveat that the current numbers are perfect-information upper bounds. If the authors add that, the paper is publishable as a modeling and benchmark contribution.","headline":"A solid integrated economic MPC formulation for price-responsive households, with ideal-planning savings that need a forecast-error robustness test before the quantitative claims can be taken at face value.","tokens_in":11468,"tokens_out":1726,"would_cite":true,"duration_ms":19337,"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":"Widening the comfort bounds an occupant allows in a smart home shifts roughly 7–10 percentage points of annual electricity use into low-price periods and cuts annual electricity bills by 16–34%, the paper contends.","keywords":["model predictive control","price-responsive loads","smart buildings","smart appliances","comfort constraints","building thermal dynamics","demand-side flexibility","economic MPC"],"falsifier":"Run the same economic MPC controller for a year in a real instrumented household (or a high-fidelity simulation with realistic forecast errors), alternating between tight and wide comfort bounds under the same tariff, and compare the annual share of electricity consumed in low-price periods. If the share does not increase by roughly 7–10 percentage points, or the annual cost does not fall by near 16–34%, the paper's central claim fails.","tokens_in":10529,"feed_emoji":"⚡","tokens_out":6915,"duration_ms":66922,"temperature":0.7,"pith_summary":"The paper sets out to show that the price responsiveness of a smart building is governed less by the heating hardware than by the comfort bounds the occupants allow the controller. It formulates an economic model predictive control problem for a single-zone household—heating or cooling, water heater, refrigerator, lighting, and uninterruptible appliances—and simulates a full year of 15-minute decisions using known day-ahead prices and one-day weather forecasts. In these simulations, widening the acceptable temperature range from zero to a few degrees raises the share of annual electricity used in low-price periods from 54.5–57.2% to 64.1–64.7% and lowers annual electricity cost by 16–34% depending on the heating system. The paper also finds that making the comfort bounds price-dependent reduces occupant discomfort relative to constant bounds, but at a higher electricity cost.","feed_headline":"Wider comfort bounds shift power to cheap hours and cut bills 16–34%","feed_subtitle":"The paper's control model shows comfort settings alone, with no new hardware, create the price response.","key_machinery":"The central object is an economic model predictive control (MPC) formulation: a mixed-integer linear program that, at each 15-minute step, minimizes electricity cost plus penalty terms for comfort violations over a one-day look-ahead horizon. The building is represented by a five-state linear thermal model—room air, floor, water in the floor-heating pipes, refrigerator chamber, and water-heater temperature—with external disturbances for ambient temperature, occupancy, and hot-water demand. Comfort enters as user-defined, time-varying upper and lower bounds on the temperatures and on light level, with slack variables penalized in the objective, so the controller is allowed to violate bounds at a price. Uninterruptible loads such as the washing machine and dishwasher are scheduled by binary variables with consecutive cycle constraints and phase-specific power draws. This machinery does the work of shifting consumption to low-price periods by preheating or precooling the building's thermal mass when electricity is cheap and deferring non-thermal loads.","core_discovery":"In the paper's own terms, the central discovery is that occupant-defined comfort constraints are the key driver of a household's price-responsive behaviour. Using the reported three flexibility cases, the annual cost falls from €103.5 to €68.4 for a water-based floor heater (a 34% saving) and from €107.0 to €76.1 for an HVAC system (a 29% saving) when comfort bounds widen from no flexibility to extra flexibility under price-independent bounds. At the same time, the share of building consumption in low-price periods rises from 54.5–57.2% to 64.1–64.7% for both systems. A separate result is that a smaller heat-transfer coefficient (better insulation) makes HVAC-equipped households more price-responsive, whereas floor-heated households become more price-responsive when less insulated; the latter is attributed to the slow thermal dynamics of the floor mass.","pith_inferences":["If the simulated 7–10 percentage point shift in low-price consumption survives real forecast error, aggregating many such homes could flatten the daily price curve and reduce peak generation needs; this is an extension the paper does not test.","The perfect-forecast, linear-model idealization means the reported savings are likely an upper bound for field operation; a testable extension is an A/B field trial with real day-ahead prices and forecast updates.","The sensitivity to the room-ambient heat-transfer coefficient suggests that insulation level and heating-system choice should be co-optimized with comfort settings under a given tariff; this joint design problem is implicit in the results but not solved here.","A natural next step not explored in the paper is to compare this comfort-bound flexibility against battery storage on a cost-per-flexibility basis for the same household."],"forward_implications":["A household can achieve substantial price responsiveness without adding batteries, solar panels, or any new equipment; the only change is widening the allowed comfort band in the controller.","The same MPC formulation handles thermostatic and non-thermostatic loads jointly, so dishwasher, dryer, and water-heater schedules shift together with heating and cooling to cheap periods.","Price-dependent comfort bounds offer a direct trade-off: occupants who care about staying near the reference temperature get less discomfort with PD-CB, but pay more than they would with uniformly wide, price-independent bounds.","The interaction between insulation and heating type matters: for fast HVAC systems, better insulation increases price responsiveness, while for slow water-based floor heating, worse insulation increases it, at higher annual cost.","Annual simulations with 15-minute decisions and integer appliance scheduling are computationally tractable (resolved in 16–60 minutes per case on one CPU), so the controller could be re-run daily in practice."],"supporting_citations":[{"why":"Supplies the base single-zone household with water-based floor heating and the economic MPC approach the paper extends.","marker":"[8]"},{"why":"Provides the state-space matrices, ambient temperature, solar radiation, electricity prices, and occupancy schedules used in the simulations.","marker":"[18]"},{"why":"Supplies the grey-box refrigerator model and its thermal parameters.","marker":"[5]"},{"why":"Supplies the domestic water heater model and demand scheduling used for the water-heater state.","marker":"[6]"},{"why":"Provides the lighting constraint formulation and the building-with-appliances modeling context.","marker":"[13]"},{"why":"Supplies residential load simulator data used for appliance and refrigerator parameters.","marker":"[15]"},{"why":"Provides the phase-wise cycle power data for the uninterruptible appliances.","marker":"[19]"}],"fun_headline_variants":[],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the controller's one-day look-ahead planning on a perfect five-state linear thermal model with known prices, weather, occupancy, and hot-water demand predicts how the real building and its occupants respond; if forecasts are imperfect or the model omits solar gain, humidity, and variable heat-pump efficiency, the simulated savings and price-shift will not be realized.","fun_headline_variants_meta":{"error":"Client error '402 Payment Required' for url 'https://api.deepseek.com/chat/completions'\nFor more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402"},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:52:52.682591+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same economic MPC controller for a year in a real instrumented household (or a high-fidelity simulation with realistic forecast errors), alternating between tight and wide comfort bounds under the same tariff, and compare the annual share of electricity consumed in low-price periods. If the share does not increase by roughly 7–10 percentage points, or the annual cost does not fall by near 16–34%, the paper's central claim fails.","supporting_citations":[{"cited_title":"Eco- nomic model predictive control for building climate contro l in a smart grid,","cited_arxiv_id":null,"evidence_quote":"Supplies the base single-zone household with water-based floor heating and the economic MPC approach the paper extends."},{"cited_title":"Ho w can smart buildings be price-responsive? – Auxiliary data","cited_arxiv_id":null,"evidence_quote":"Provides the state-space matrices, ambient temperature, solar radiation, electricity prices, and occupancy schedules used in the simulations."},{"cited_title":"Grey-box modeling for system identification of household r efrigerators: A step toward smart appliances,","cited_arxiv_id":null,"evidence_quote":"Supplies the grey-box refrigerator model and its thermal parameters."},{"cited_title":"Scheduling of domestic water heater power demand for maxim izing PV self-consumption using model predictive control,","cited_arxiv_id":null,"evidence_quote":"Supplies the domestic water heater model and demand scheduling used for the water-heater state."},{"cited_title":"De- centralized coordination of a building manager and an elect ric vehicle aggregator,","cited_arxiv_id":null,"evidence_quote":"Provides the lighting constraint formulation and the building-with-appliances modeling context."},{"cited_title":"Smart residential load simulator for energy man- agement in smart grids,","cited_arxiv_id":null,"evidence_quote":"Supplies residential load simulator data used for appliance and refrigerator parameters."},{"cited_title":"Synergy potential of smart a ppliances. Report of the Smart-A project,","cited_arxiv_id":null,"evidence_quote":"Provides the phase-wise cycle power data for the uninterruptible appliances."}],"review_version":1}