{"id":"4f8feaf5-1c6c-4f36-b464-7c453220b1fe","arxiv_id":"2511.18267","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"An off-the-shelf residential heat pump runs on 350 V DC with performance equivalent to AC, and modeled DC-nanogrid operation cuts annual bills by 12.5–16.7%.","lead":"This paper tests whether a normal AC-powered residential heat pump can be fed directly with 350 V DC after a small retrofit, in both a lab and an occupied house. It finds the heat pump performs about the same on DC, and a simulation suggests wiring the house as a DC solar nanogrid could cut annual electricity bills by 12.5% to 16.7%.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Converter part-load curves are the least-supported input; the 12.5–16.7% annual savings estimate is not robust until those curves are replaced with measured or manufacturer-specific data.","rationale":"The reader's weakest assumption identifies the same load-bearing concern I find: the system-level savings are produced by stylized converter part-load curves whose shape is not measured and whose sensitivity is unreported. This is not a manufactured objection—the paper's own Section 6.2 limitation states that a single stylized curve was used for all converters and that tailoring full curves would be more sophisticated. Because the model holds heat pump demand identical between AC and DC, every dollar of the 12.5–16.7% savings is attributable to modeled conversion losses, making the converter curves the sole mechanism for the headline result. The proposed test is cheap and decisive: re-run the published code with realistic per-converter part-load data. The device-level claim, by contrast, is reasonably supported: AHRI laboratory tests show capacity and COP within measurement uncertainty, the field t-test (p=0.18) does not reject equality, and the authors disclose the main confounds (one-year-apart AC data, warmer AC weather, possible degradation). Those limitations weaken but do not overturn the experimental contribution. The verdict should remain CONDITIONAL: the hardware demonstration is credible, but the annual bill savings should be treated as conditional on the converter part-load model being representative. I agree with the reader rather than escalating to REJECT because there is no internal inconsistency in the experiments and the modeling limitation is explicitly acknowledged; it is an addressable condition, not a fatal flaw.","tokens_in":17115,"tokens_out":5157,"duration_ms":61584,"concrete_test":"Using the public GitHub repository (github.com/AaronFarha/dchp), replace Fig. 13's single stylized part-load curve with manufacturer-specific efficiency-vs-load data for each converter—or, at minimum, with a conservative curve that is 5 points lower below 30% load—and re-run the annual bill comparison. Report the resulting AC vs DC bills for the after-market and ideal-DC configurations. If the 12.5%/16.7% savings fall below roughly 5%, or if the ordering between configurations changes, the system-level claim is not supported by the current evidence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central system-level claim is that measured device-level parity plus modeled conversion-loss reductions yield 12.5–16.7% annual bill savings. In the simulation, the heat pump's electricity demand is identical under AC and DC, so all savings come from the converter efficiency model. Section 5.2 states that only the peak of each converter's part-load curve was tuned to a manufacturer specification sheet, and Fig. 13 uses a single stylized curve shape for every converter. Section 6.2 explicitly concedes that a more sophisticated approach would tailor each converter's full part-load efficiency curve. This matters because real converters, especially DC-DC stages and inverters, often have substantially lower efficiency at 10–30% load, and a house operates in that low-load region for most of the year. The paper reports no sensitivity analysis for these curves. If actual part-load efficiencies are even 5–10 points lower in the low-load region, the modeled conversion-loss savings shrink directly and could fall below the 10-year simple-payback threshold (612 USD capital cost ceiling) that the authors themselves use. The full-retail net-metering assumption (0.14 USD/kWh) similarly inflates the summer export credits shown in Fig. 14, but the part-load curves are the more load-bearing unsupported input because they are the sole source of the savings in the model.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports laboratory and field experiments in which an off-the-shelf variable-speed residential heat pump designed for 230 V AC is retrofitted to operate on 350 V DC. AHRI Standard 210/240 psychrometric-chamber tests show that thermal capacity and COP differ by at most about 6% between AC and DC operation, comparable to measurement uncertainty. A one-month field test in an occupied house, comparing weather-normalized daily input power with AC data collected one year earlier, yields p = 0.18 in a two-sample t-test, which the authors interpret as no statistically significant performance difference. These device-level results are then used to parameterize a year-long nanogrid simulation with PV, battery, and the heat pump, predicting annual electricity-bill savings of 12.5% for the after-market DC retrofit and 16.7% for an idealized DC-motor heat pump, attributed to reduced AC/DC conversion losses.","tokens_in":17469,"tokens_out":6054,"duration_ms":58056,"significance":"The experimental contribution is genuinely valuable: it provides the first laboratory heating-mode results and the first field results for a retrofit DC heat pump, with careful attention to AHRI test conditions, energy-balance closure to within 6%, documented sensor uncertainties, and a formal hypothesis test. The public availability of data and code is a further strength. The device-level parity claim is plausible and well supported. However, the headline system-level savings are not yet robust: the simulation's savings come entirely from a stylized converter part-load model whose curve shapes are not empirically supported, and the field parity evidence is weakened by non-concurrent AC data and a stated risk of DC-induced degradation. If the modeling uncertainties are addressed, the paper would be a strong experimental and modeling contribution to the DC-nanogrid literature.","major_comments":[{"comment":"The annual savings estimates (12.5% and 16.7%) are outputs of the converter part-load model, since the same historical heat-pump demand is used for AC and DC in the simulation (§5.1). The text states that only the peak of each converter's part-load curve was tuned to manufacturer data and that Fig. 13 uses a single stylized curve shape for all converters; §6.2 explicitly concedes that tailored full part-load curves are needed. Real converters often lose efficiency disproportionately at the low loads that dominate residential operation. No sensitivity analysis is provided. If actual part-load efficiencies are even a few points below the modeled curves in the low-load region, the conversion-loss savings shrink and the 612 USD ten-year-payback ceiling used in §6.1 may no longer be met. Please replace the stylized curves with measured or manufacturer-specific curves, or add a conservative se","section":"§5.2, Fig. 13, §6.2"},{"comment":"The field-equivalence result is load-bearing for the claim that the nanogrid savings stem from conversion-loss reductions rather than from heat-pump performance differences. The AC reference data were collected one year earlier in warmer weather (mean ΔT 18.7 vs 21.1 °C), and the paper itself notes possible performance degradation under DC operation. The t-test normalizes by Eq. (3) with c = 8 °C fixed, but no sensitivity to c is reported. The results paragraph also contains an internal inconsistency: p = 0.18 is reported, followed by “Because 0.23 > 0.05”. Please correct the typo and strengthen the analysis with a sensitivity study over c and/or an equivalence/non-inferiority test; otherwise the conclusion of “no statistically significant difference” may overstate the evidence for parity.","section":"§4.2, Eq. (3)-(5)"},{"comment":"The bill savings include credits for exporting surplus solar at the full retail rate of 0.14 USD/kWh. Under net-metering policies that compensate exports at avoided cost or wholesale rates, the summer negative bills in Fig. 14 would shrink, and so would the annual savings. Since this assumption, together with the converter part-load curves, drives the headline savings, it should be reported as a conditional assumption and tested with a sensitivity case, e.g., export price = 0 or a typical export tariff.","section":"§5.1, Fig. 14"}],"minor_comments":[{"comment":"The text reports R² values of 0.895 for AC and 0.931 for DC, then says “The AC fit likely has a higher R² value”; this is contradicted by the reported numbers and the following reasoning should be revised.","section":"§4.2, near Fig. 10"},{"comment":"The sentence “For intermediate temperature differences, the heat pump appears to draw slightly more power on AC than on DC” appears to be a typo for “more power on DC than on AC”, since the preceding sentence already states that DC draws less power at low and high temperature differences.","section":"§4.2, paragraph after Fig. 10"},{"comment":"The symbol b is used both as a scalar battery power capacity (b = 12.5 kW) and as a time-varying battery power b(k) in Eq. (8) and Eq. (9). Please use distinct symbols to avoid ambiguity.","section":"§5.1, Eq. (8)-(9)"},{"comment":"The claim of “the first modeling results informed by real-world data from an occupied test house” is strong; earlier work by Vossos et al. and others also used measured load data. Consider softening or documenting the exact basis for the claim.","section":"§2.4 and §7"}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here’s my read on arXiv:2511.18267. The device-level claim is solid: an off-the-shelf variable-speed heat pump, fed 350 V DC instead of 230 V AC after a modest outdoor-unit retrofit, delivers equivalent capacity and COP in AHRI-standard chamber tests (heating and cooling), and the field comparison in an occupied house shows no statistically significant difference (p=0.18). That is genuinely new — prior work on this same pump covered only steady cooling at full speed — and the authors have done the experiments carefully: energy balances within 6%, documented sensor uncertainties, and a weather-normalized Welch t-test. They also ship data and code. Credit where due.\n\nThe soft spot is not the heat pump; it’s the system-level savings estimate. The 12.5–16.7% annual bill reduction in the DC nanogrid is a model output in which the heat pump’s electricity demand is identical on AC and DC, so all savings come from the modeled converter efficiencies. Section 5.2 says only the peak of each converter’s part-load curve was tuned to a manufacturer spec sheet, and Fig. 13 uses a single stylized curve shape for every converter. Section 6.2 concedes a more tailored approach is needed. Real converters often sag more at 10–30% load, and that is where a house lives most of the year. No sensitivity analysis is reported. The full-retail buy-back at 0.14 USD/kWh also sweetens summer exports. The authors flag these limitations, and the paper is honest about them, but they are not quantified, so the annual savings number should be treated as an upper-end estimate, not a robust prediction. The paper’s own simple payback ceiling of 612 USD makes clear how much rests on that number.\n\nOne minor typo: the text says “Because 0.23>0.05” right after reporting p=0.18. Not a substantive problem.\n\nOverall: this is a serious paper. The experimental contribution will be the reference point for DC heat-pump retrofits for a while. The modeling section needs revision — real part-load curves and a sensitivity analysis — before the savings claim should be taken at face value. But the device-level finding is the main event and it holds up.\n\nFor you: worth reading if you care about DC distribution or heat pump integration. I’d cite it. And yes, if this came to my desk, I’d send it to peer review — the experiments deserve referee time; the model may get revised but that’s what review is for.","headline":"First hardware evidence that a residential heat pump can run on a 350 V DC bus without measurable performance loss, though the 12.5–16.7% bill-savings number is a model output resting on stylized converter part-load curves.","tokens_in":17990,"tokens_out":1869,"duration_ms":16813},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"An off-the-shelf 230 V AC heat pump can be run on a 350 V DC bus after a minor outdoor-unit retrofit, with no significant change in capacity or efficiency, and a DC nanogrid built around it could cut annual electric bills by 12.5–16.7%.","keywords":["DC nanogrid","heat pump retrofit","direct current distribution","solar photovoltaics","battery storage","variable-speed heat pump","experimental validation","residential electricity bills"],"falsifier":"Measure the actual efficiency of each converter in the chain—PV inverter, battery DC-DC converter, and heat-pump rectifier/inverter—across the load distribution implied by the 2024 data, then recompute the annual bill. If the measured annual cycle-weighted converter losses differ materially from the stylized curves, the predicted 12.5–16.7% savings would not hold; alternatively, run the same house's whole nanogrid for a full year in AC and then in DC and compare the actual bills.","tokens_in":17007,"feed_emoji":"⚡","tokens_out":3540,"duration_ms":36502,"temperature":0.7,"pith_summary":"The paper tries to establish that a normal variable-speed residential heat pump, designed for 230 V AC, can be powered directly from a 350 V DC bus after a small retrofit, with no meaningful loss in heating or cooling performance. Laboratory tests across standard rating conditions found AC and DC thermal capacity agreed within about 6%, and COP differences were within measurement uncertainty. A one-month winter field test in an occupied house confirmed the device-level parity: weather-normalized input power was statistically indistinguishable between AC and DC. The authors then use the measured heat-pump load in a year-long model of a solar-plus-battery nanogrid, showing that DC distribution lowers the test house's annual electric bill by 12.5% with the retrofit and 16.7% with a heat pump designed for DC. The savings come from removing AC/DC conversions in the rest of the nanogrid, not from the heat pump itself, and the paper gives an experimental anchor to the idea that DC distribution can reduce residential electricity costs.","feed_headline":"DC power can cut a test house's electric bill by up to 17%","feed_subtitle":"An off-the-shelf heat pump kept its capacity and efficiency on 350 V DC, so the savings come from removing AC/DC conversions.","key_machinery":"The carrying object is the DC nanogrid bus and the converter chain around it. In the AC configuration, solar and battery each require inverters, and the heat pump internally rectifies AC back to DC before its own inverter drives the compressor. The retrofit feeds 350 V DC directly into the heat pump's variable-speed drive, bypassing the onboard rectifier. The quantitative engine is a set of converter part-load efficiency curves—one stylized curve shape, tuned at its peak to manufacturer spec-sheet values for each converter type—combined with a battery control logic that prioritizes solar for the heat pump first, then battery charging, and a full year of measured heat-pump and rest-of-house l","core_discovery":"The paper's central claim is that a residential variable-speed heat pump sold for conventional AC service can be retrofitted to run on a 350 V DC bus with minimal hardware changes and with no statistically significant change in capacity or coefficient of performance. In steady-state laboratory tests spanning cooling and heating conditions, the heat pump delivered essentially the same thermal capacity on DC as on AC, with COP differences within the measurement uncertainty (about 6%). A winter field test in an occupied 208 m² house, comparing weather-normalized input power on AC versus DC, found the difference statistically insignificant at p = 0.18. The paper then argues that the real benefit","pith_inferences":["The 12.5–16.7% savings are sensitive to the assumed converter part-load curves and to the assumption that surplus solar is sold at full retail price; under real-world export tariffs or less efficient partial-load converters, the savings would be smaller.","The paper's equivalence result applies to a heat pump whose compressor still uses an internal inverter; a heat pump with a true DC-driven compressor might show additional gains or different reliability characteristics that the current retrofit cannot reveal.","The same retrofit logic extends naturally to other natively-DC residential loads, such as EV chargers and LED lighting, so the system-level savings likely grow as more loads are added to the DC bus.","A more decisive field test would switch the same heat pump between AC and DC within the same season, avoiding the year-apart weather and equipment-aging confound that the current field comparison carries."],"forward_implications":["Existing AC heat pumps can be retrofitted to run on DC without sacrificing capacity or efficiency, so a DC distribution upgrade does not require replacing the largest residential load.","The 12.5% and 16.7% annual bill savings in the modeled house come from eliminating inverter/rectifier losses and improving battery round-trip efficiency, not from the heat pump operating better on DC.","Given a ten-year simple payback ceiling, the highest justifiable retrofit cost for this house is about $612, meaning DC retrofits are not economically attractive today for single-family homes on energy savings alone.","The device-level parity shown in the lab and field justifies moving to full system-level field trials of DC nanogrids with heat pumps, batteries, and solar arrays.","A heat pump designed natively for DC, without the internal rectifier and inverter stages, offers a larger modeled saving (16.7% vs. 12.5%) than the after-market retrofit."],"fun_headline_variants":["DC retrofit for heat pumps: bill slashed up to 17%","Heat pump runs on DC with no performance loss","Simulated DC nanogrid cuts electric bills by 17%","Retrofit heat pump to DC, same power, savings","DC power could cut home energy bills by 17%"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The bill savings rest on modeled part-load efficiency curves for the converters, whose peak values come from specification sheets and whose shape is the same stylized curve for every converter; if real converters are less efficient at the partial loads where the house actually operates, the 12.5–16.7% savings shrink.","fun_headline_variants_meta":{"raw":{"variants":["DC retrofit for heat pumps: bill slashed up to 17%","Heat pump runs on DC with no performance loss","Simulated DC nanogrid cuts electric bills by 17%","Retrofit heat pump to DC, same power, savings","DC power could cut home energy bills by 17%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000178,"raw_usage":{"total_tokens":1128,"prompt_tokens":733,"completion_tokens":395,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":477,"completion_tokens_details":{"reasoning_tokens":310}},"tokens_in":477,"tokens_out":395,"duration_ms":5433,"temperature":1.0,"reasoning_tokens":310,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T20:46:24.500294+00:00","model_set":{"reader":"deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.3.0-alltime-deepseek","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":"Measure the actual efficiency of each converter in the chain—PV inverter, battery DC-DC converter, and heat-pump rectifier/inverter—across the load distribution implied by the 2024 data, then recompute the annual bill. If the measured annual cycle-weighted converter losses differ materially from the stylized curves, the predicted 12.5–16.7% savings would not hold; alternatively, run the same house's whole nanogrid for a full year in AC and then in DC and compare the actual bills.","supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}