{"id":"ffa9b0c3-e86e-49a7-89ca-029827f3f40e","arxiv_id":"2607.28280","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Adaptive demand-driven PCM control smooths campus district-heating peaks better than rule-based charging, with best results near 80°C phase-change temperature and conductivity above ~2 W/(m·K), but payback remains ~25 years.","lead":"A campus district-heating simulation shows adaptive demand-driven control of PCM storage cuts peak city-network heat draw about 5–11% and avoids the secondary charging peaks that fixed schedules create. The result matters for operators facing peak tariffs and waste-heat recovery, though payback stays ~25 years under current costs.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.5","headline":"ADD peak claims rest on an unvalidated plant model and a daily threshold that uses same-day max demand, so the 5.3%/11.3% figures may not transfer.","rationale":"The reader correctly locates the soft spot: unvalidated integrated dynamics plus peak definition. I sharpen one operational detail that makes the claim least secure for transfer—the same-day Q_daily,max,d inside the ADD law (Eq. 23)—because that is what makes the controller look non-anticipative in the paper while still using information a real plant would not have at decision time. The qualitative ADD-vs-RBC contrast (no new off-peak peaks, comfort and COP largely preserved) is internally coherent under the stated model and does not require REJECT. CONDITIONAL remains the right verdict: accept as model-based evidence if limited to “in this simulation,” with the causal-threshold check (and ideally code/data) needed before treating 5.3%/11.3% or 25.3-year payback as plant-ready. No change to the reader’s overall posture; agreement is agree on the load-bearing issue.","tokens_in":21814,"tokens_out":713,"duration_ms":12671,"concrete_test":"Re-run the 31-day winter case and the annualized Table 5 case with a causal threshold only: replace Q_daily,max,d in Eq. 23 by the previous day’s max (or a one-day-ahead forecast with realistic error). If peak-load reduction falls materially below ~5% (winter) / ~11% (annual) or secondary charging peaks reappear in the Fig. 8-style trace, the transferable magnitude of the strongest claim is not supported.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim (ADD smooths city DH extraction and delivers ~5.3% winter / up to 11.3% annualized peak reduction vs baseline, without RBC-style charging peaks) is entirely simulation-based. The end-to-end Python plant—effective-heat-capacity PCM PDE (Eqs. 5–6), semi-empirical HP map (Eq. 3, Fig. 3), lumped RC building/radiators (Eqs. 11–17), and energy balances (Eqs. 18–22)—has no integrated experimental validation against the Trondheim campus; only component literature is cited. Peak metrics further depend on averaging the three highest DH peaks (Eq. 21) and on a daily threshold Q_th,d = (1-α_sh) Q_daily,max,d (Eq. 23, α_sh=15%) that uses the same day’s reference maximum. In a true online controller that quantity is not known a priori; if the implemented threshold is only a forecast or a lagged statistic, headroom-limited charging (Eqs. 24, 27) and overload-driven discharge (Eqs. 25, 28) can mis-time, re-create secondary peaks, or cut the realized shift well below the reported 5.3%/11.3%. Model bias in discharge capability, charging headroom, or this oracle threshold would therefore move the headline numbers and the 25.3-year payback without overturning the qualitative ADD-vs-RBC story inside the model.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"This paper develops a dynamic Python simulation of a campus-scale two-stage district heating (DH) network in Trondheim with heat-pump recovery of data-centre waste heat and a shell-and-tube PCM storage tank. It compares a baseline weather-compensated system, a schedule-based rule-based PCM controller (RBC), and a proposed adaptive demand-driven (ADD) controller that charges only under network headroom and discharges only under overload relative to a daily peak threshold. Performance is assessed via peak DH extraction (average of the three highest peaks), operating cost under energy-plus-demand tariffs, indoor comfort, and HP COP, with sensitivity on PCM phase-change temperature, latent heat, and thermal conductivity, plus a CAPEX/payback analysis. The central claim is that ADD smooths city-network heat extraction, avoids RBC charging peaks, and delivers about 5.3% peak reduction in a 31-day winter window and up to 11.3% annualized peak reduction with ~1.4% operating-cost savings, at a simple payback of 25.3 years; recommended PCM settings are Tm ≈ 80 °C and k ≳ 2 W/(m·K).","tokens_in":22172,"tokens_out":1807,"duration_ms":38049,"significance":"If the quantitative peak-shaving and cost results transfer beyond the unvalidated plant model, the work is a useful systems-level contribution: it couples PCM thermophysics with a practical (non-MPC) control layer and shows that control design can dominate storage hardware for demand-charge reduction. The explicit baseline/RBC/ADD comparison under shared weather and comfort checks, the temperature-distribution diagnostics that explain Tm and k trends, and the frank 25.3-year payback are strengths. The ADD logic (headroom-limited charge, overload-limited discharge) is a clear, implementable idea for DH operators. Significance is currently limited by the lack of integrated model validation and by an oracle daily threshold that uses same-day maximum demand, so the headline 5.3%/11.3% figures should be treated as model-internal until those issues are addressed.","major_comments":[{"comment":"The end-to-end plant (effective-heat-capacity PCM PDE Eqs. 5–6, semi-empirical HP map Eq. 3/Fig. 3, lumped RC building/radiators Eqs. 11–17, energy balances Eqs. 18–22) has no integrated experimental or operational validation against the Trondheim campus. Only component-level literature is cited. Because the central claims (5.3% winter peak cut in Fig. 9; 11.3% annualized in Table 5; 25.3-year SPP) are entirely simulation deltas in Q_DH and cost, the manuscript needs either (i) validation of key trajectories (substation heat rates, HP COP, storage inlet/outlet temperatures) against campus data or a published benchmark, or (ii) a structured uncertainty/bias analysis showing how peak and payback move under plausible model error. Without this, transferability of the quantitative results remains unestablished.","section":"§2.3, Eqs. 3–22; §4.1; Table 5"},{"comment":"The ADD daily threshold is Q_th,d = (1−α_sh) Q_daily,max,d with α_sh = 15% (Eq. 23, Table 3). Q_daily,max,d is the same day’s reference maximum and is therefore not available to a causal online controller. Headroom charging (Eqs. 24, 27) and overload discharge (Eqs. 25, 28) can look artificially well-timed under this oracle. Please replace or stress-test with a causal threshold (e.g., previous-day peak, rolling percentile, or day-ahead forecast with error) and report how the 5.3%/11.3% peak reductions and secondary-peak avoidance change. If the oracle is retained only as an upper bound, state that explicitly and demote the online-performance claim.","section":"§3.3, Eqs. 23–28; Table 3"},{"comment":"Peak reduction is reported as ~5.3% for the 31-day winter comparison (Fig. 9, abstract) but 11.3% annualized (Table 5), with annual peak loads 98.3 → 87.2 MW. The annualization method is not described (scaling of the 31-day window? full-year weather? different peak definition?). Clarify the procedure, reconcile the two figures, and state which metric supports the techno-economic cash flows (S_ann, SPP 25.3 years). Inconsistent peak metrics undermine the load-bearing economic claim.","section":"§4.1.2 Fig. 9; §4.3 Table 5; Eqs. 21–22, 29–36"},{"comment":"Storage sizing uses a target excess energy above a threshold (Eqs. 1–2) with utilization factor η_u, while ADD is tuned with α_sh = 15% and 2 MW charge/discharge caps (Table 3), yet realized winter peak reduction is only 5.3%. Please report achieved shifted energy versus E_target, average state-of-charge / latent-heat utilization under ADD, and whether under-performance is due to thermal conductivity, temperature mismatch, or the headroom logic. Without utilization metrics, the sensitivity conclusions on Tm and k (Figs. 10–15) are harder to interpret at system level.","section":"§2.2 Eqs. 1–2; §3.3 Table 3; §4.2"}],"minor_comments":[{"comment":"Nomenclature lists DPP (discounted payback period) but the main text and Table 5 emphasize SPP; Fig. 16 uses DPP. Define DPP explicitly alongside SPP and state the discount rate(s) used in Fig. 16.","section":"Nomenclature; §3.4; Fig. 16"},{"comment":"Fig. 8 encoding appears corrupted in the manuscript source (unicode private-use sequences), which harms readability of the key load-profile comparison. Ensure a clean vector figure in revision.","section":"Fig. 8"},{"comment":"Abstract says “up to 5.3% peak-load reduction” while conclusions lead with 11.3% annualized; align abstract, §4, and §5 on which figure is primary.","section":"Abstract; §5"},{"comment":"Table 1 lists building heated volume 1,054,600 m³ and very large envelope capacitance; briefly justify that these represent the full campus aggregate rather than a single building, since §2.3.3 says “a representative building is considered as the campus heating demand.”","section":"Table 1; §2.3.3"},{"comment":"HP fit (Eq. 3) coefficients a–g are not tabulated; providing them (or a data/code supplement) would aid reproducibility.","section":"§2.3.1 Eq. 3"},{"comment":"Minor language/typos: “Latent heat thermalenergy storage”, “peak load reduction” vs “peak-load”, and arXiv dates in references (e.g., 2026) should be checked for consistency before publication.","section":"Abstract; References"}],"recommendation":"major_revision","confidential_remarks":"The qualitative ADD-vs-RBC story (avoid charging peaks via headroom limits) is plausible and worth publishing after revision; the skeptic’s oracle-threshold and validation concerns are real and load-bearing for the quantitative headlines, not pedantic. I would not reject on novelty grounds—the integrated PCM+HP+ADD campus case is a reasonable contribution for an energy-systems journal—but I would not accept the 5.3%/11.3%/25.3-year numbers without causal threshold tests and clearer annualization. Scope fit for eess.SY / applied energy venues is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful bit here is not another PCM tank paper. It is a clean head-to-head, on one campus DH model with HP waste-heat recovery, of fixed-schedule RBC versus a simple adaptive rule that charges only into network headroom and discharges only on overload. Inside that model the qualitative story holds: RBC shaves early peaks and creates charging spikes; ADD flattens the city-side draw (~5.3% in the winter window, ~11.3% annualized in their table) without wrecking comfort or HP COP. The PCM sensitivity is also done properly—Tm around 70–80°C and k above ~2 W/(m·K) matter; latent heat barely moves the needle in their range—and the temperature-distribution plots actually explain why.\n\nWhat is new is the combination and the system-level bookkeeping, not the ingredients. Lee, Rinaldi, Thomson, Calise and others already covered PCM peak shaving, HP-assisted storage, and smarter TES control. The authors are honest about the 25.3-year payback and the need for tariff/incentive change; that honesty helps more than the usual deployment rhetoric.\n\nSoft spots, in proportion. The end-to-end Python plant (effective-capacity PCM PDE, fitted HP map, lumped RC building) has no integrated validation against Trondheim—only component literature. So treat 5.3%/11.3% and the payback as model outcomes, not plant facts. The stress-test point on the daily threshold is real: Q_th,d uses the same day’s max demand, which an online controller does not know. If that becomes a forecast or lag, timing and realized shift can degrade; the paper should say so. CAPEX is factor-based, free parameters (α_sh, power caps, η_u, cost factors) are many, and code/data are not released. None of that overturns the internal ADD-vs-RBC comparison.\n\nWho it is for: people designing TES control and PCM selection for DH under peak tariffs. Worth a serious referee. I would send it to review with requests to flag the oracle threshold, separate winter vs annualized peaks clearly, and release enough model detail to reproduce the trajectories. Engage if you work peak flexibility or LHTES control; skim the sensitivity figures if you only need the property band.","headline":"Solid campus-scale sim showing headroom/overload PCM control beats schedule RBC on peak shape; numbers rest on an unvalidated model and a same-day peak threshold.","tokens_in":22892,"tokens_out":586,"would_cite":false,"duration_ms":18103,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"Adaptive demand-driven control lets PCM storage cut district-heating peaks without creating new ones, but payback stays long.","keywords":["district heating","thermal energy storage","phase change material","peak shaving","adaptive demand-driven control","techno-economic analysis"],"falsifier":"Instrument the real campus main substation and a pilot PCM tank under both schedule-based and ADD logic for one winter and check whether measured city-network peak (average of three highest) falls by roughly 5–11 percent without new off-peak spikes or comfort loss.","tokens_in":22584,"feed_emoji":"🔥","tokens_out":950,"duration_ms":19553,"temperature":0.7,"pith_summary":"District heating networks face sharp winter peaks that force oversized plant and high demand charges. This paper argues that a shell-and-tube phase-change storage tank, charged from the city network and discharged to buildings, can shift those peaks if it is controlled by real-time network headroom and overload rather than by a fixed clock. In a dynamic model of a Norwegian campus system with heat-pump waste-heat recovery, the adaptive strategy smooths city heat extraction, cuts the representative-period peak by about 5.3 percent (11.3 percent annualized), and avoids the secondary charging spikes that simple schedule-based rules create, all while holding indoor comfort. Material sensitivity points to a melting temperature near 80 °C and conductivity above roughly 2 W/(m·K) as the sweet spot. The same runs show a simple payback of about 25 years under present tariffs, so cost cuts or stronger peak incentives are still required before the approach is economically attractive.","feed_headline":"PCM storage cuts heating peaks 5–11% when control tracks demand","feed_subtitle":"Adaptive charging avoids new spikes; payback still ~25 years without cheaper materials or stronger tariffs","key_machinery":"Adaptive demand-driven (ADD) PCM control: daily peak threshold plus headroom-limited charging and overload-limited discharging (with candidate time windows retained only as soft bounds), coupled to an effective-heat-capacity shell-and-tube PCM model inside a full campus DH–heat-pump–building simulation.","core_discovery":"Under a dynamic campus district-heating model, adaptive demand-driven (ADD) control of a PCM tank—charging only when network headroom exists and discharging only when load exceeds a daily threshold—smooths city-network heat extraction and delivers roughly 5.3 percent peak reduction in a winter comparison (up to 11.3 percent annualized) without secondary charging peaks or loss of indoor comfort, outperforming fixed-schedule rule-based control; a melting point near 80 °C and conductivity above 2 W/(m·K) further improve results, yet simple payback remains about 25 years.","pith_inferences":["If dynamic network tariffs begin to reward continuous load flattening rather than only the three highest monthly peaks, the same ADD–PCM package would become far more bankable than the 25-year figure suggests.","The same headroom/overload logic could be ported to other low-temperature TES media or to multi-building clusters once local mass-flow and return-temperature measurements exist.","Because latent-heat capacity showed only marginal system-level gains once conductivity and melt temperature were adequate, future material R&D for DH may yield higher returns from heat-transfer enhancement than from enthalpy alone."],"forward_implications":["Fixed time-of-day charging of PCM storage can raise, rather than lower, the billed peak and should not be treated as a default control layer.","PCM selection for DH should prioritize melting temperature match (~70–80 °C) and conductivity ≳2 W/(m·K) over simply maximizing latent heat.","Under present demand-charge structures the storage investment alone is unlikely to pay back inside a typical project life without material-cost cuts or stronger flexibility tariffs.","Headroom- and overload-aware logic can be layered on existing weather-compensated DH controls without requiring full model-predictive infrastructure."],"fun_headline_variants":["ADD control of PCM tanks cuts DH peaks 5.3% without secondary spikes","PCM at 80°C and >2 W/mK conductivity boosts peak shave and economics","Demand-driven PCM charging smooths heat loads better than fixed rules","Adaptive PCM strategy trims winter peaks 5.3%, annualized up to 11.3%","PCM-integrated DH with ADD control: flexibility gains, 25-year payback"],"cache_read_input_tokens":128,"weakest_assumption_plain":"The unvalidated end-to-end dynamic simulation is assumed accurate enough that its peak-reduction and payback numbers would hold on the real campus plant.","fun_headline_variants_meta":{"raw":{"variants":["ADD control of PCM tanks cuts DH peaks 5.3% without secondary spikes","PCM at 80°C and >2 W/mK conductivity boosts peak shave and economics","Demand-driven PCM charging smooths heat loads better than fixed rules","Adaptive PCM strategy trims winter peaks 5.3%, annualized up to 11.3%","PCM-integrated DH with ADD control: flexibility gains, 25-year payback"]},"model":"grok-4.5","effort":"low","cost_usd":0.003718,"raw_usage":{"total_tokens":1260,"prompt_tokens":906,"num_sources_used":0,"completion_tokens":97,"cost_in_usd_ticks":37184000,"prompt_tokens_details":{"text_tokens":906,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":257,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":906,"tokens_out":97,"duration_ms":5198,"temperature":1.0,"reasoning_tokens":257,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T12:23:12.287275+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Instrument the real campus main substation and a pilot PCM tank under both schedule-based and ADD logic for one winter and check whether measured city-network peak (average of three highest) falls by roughly 5–11 percent without new off-peak spikes or comfort loss.","supporting_citations":[],"review_version":1}