May this 1.177 be stated as the contract ratio of 1.20 being achievable? The residual protecting that ratio is 275 kW. One line in the same load table can move 750 kW. The load chapter states the result and then withdraws the tone in which it could be quoted: in the baseline column — summer design condition, most permissive supply-temperature branch, liquid fraction 0.95 — facility power reaches 14,125 kW against a 14,400 kW contract basis, leaving 275 kW or 1.9 %, at a ratio of 1.177 against 12,000 kW of information-technology load [MODELED]. Set beside three uncertainty widths from the same table, the residual loses.
Modeled 1.177 Flips If Any One Assumption Goes Adverse
The rejection fan line spans 255 kW across its 315 to 570 kW range; the mechanical refrigeration line differs by 750 kW between the baseline and conservative columns; half a percentage point of uninterruptible-supply efficiency is 65 kW [SENSITIVITY]. The residual is smaller than the uncertainty of one line item and far smaller than that of another — none of the three being an independently established uncertainty, only a width inside one assumed table.
So the conclusion written from that comparison is that the 1.177 does not possess engineering robustness: the adverse end of any single assumption flips it. The author assigns the result an evidence grade, a parametric estimate built on engineering experience parameters, and rules that it may not be stated externally as a claim that 1.20 is achievable, nor spoken in the register of a demonstrated conclusion.
Expressed in the units of the thing it protects, 275 kW against 12,000 kW is 0.023 in ratio terms, and 1.177 plus 0.023 returns 1.20 [MODELED]. The whole distance between the modeled result and the contract basis is 23 thousandths. Quoting three decimal places across that distance is the mismatch the conclusion objects to — a characterisation of the presentation, not a recomputation.
The 14,400 kW contract basis and the 15,000 kW interconnection cap are two independent limits. They must not be merged into a single question about whether the plant is over 15 MW. Uncertainty of plus or minus 750 kW is already larger than the 600 kW between those two caps.
Four Closable Interfaces Drive the Conservative Overshoot
The same section decomposes the conservative-column overshoot rather than reporting it, which turns a binary question into a work list. Mechanical refrigeration moving one supply-temperature branch contributes 750 kW, 44.5 %. Secondary pumps moving from 105 to 385 kW contribute 280 kW, 16.6 %. Rejection fans moving from 35 to 55 kW per thermal megawatt contribute 260 kW, 15.4 %. Primary pumps moving from 200 to 400 kW contribute 200 kW, 11.9 %. Reserve moving from 3 % to 5 % contributes 133 kW, 7.9 %; the air-side terminal at a lower liquid fraction contributes 70 kW, 4.2 % [MODELED].
Each increment is a modelled difference produced by one assumption move, not an observed sensitivity. The author's reading is that the top four are 88.4 % of the overshoot, that every one is a technical interface closable before design freeze, and that not one is an uncontrollable external factor.
That conservative column is a forbidden operating region, not a conservative design to publish. It also hits a third limit: generation N+1. None of the three mitigations that work on the contract figure works there.
The Contract Cap Binds Before the Physical Supply Cap
The residual against two separate limits finds the commercial one binds earlier: 642 kW against the physical supply ceiling once battery recharge is included, against 275 kW on the contract basis. Exceeding the ratio can be mitigated by negotiation, by clarifying which weather condition and measurement window the contract figure refers to, or by operational optimisation — none of which requires changing hardware. Only a slide towards the conservative column reaches a boundary that forces the electrical supply arrangement to be redone. Resources should therefore close the top four interfaces rather than apply for more capacity.
Battery recharge does not enter the steady ratio numerator. It must enter the 15,000 kW demand check. The exclusion has to be written into the measurement clause, or a summer restore-plus-charge hour is a default.
Do Not Subtract Recharge Twice From the Two Margins
Shared engineering memory records a steady-state margin of 609 kW against the physical ceiling, computed from a facility power of 14,391 kW, with its own line-by-line load table noted as still to be verified. The load chapter's revision reports 875 kW against the same ceiling. The design did not improve: facility power was recomputed line by line from 14,391 to 14,125 kW, a net reduction of 266 kW, while the accounting became more complete — the air-side terminal, the adiabatic auxiliaries and the battery recharge term appear explicitly for the first time.
The same conclusion bars an arithmetic a reader would find natural. The 875 kW steady-state margin and the 642 kW margin including recharge are two accounting bases for one column, differing exactly by the 233 kW recharge term; subtracting the recharge from 642 to obtain 409 kW deducts the same term twice, and that value is now barred from capacity checks, generator sizing, rejection calculations, ratio calculations and investment estimates.
The Independent Check Proved Arithmetic, Not the Plant
The independent check record filed with the distribution-unit review classifies itself in its first field as arithmetic and supplied-network-snapshot checks only, not equipment validation, and its last field records that no independent solver was available [MODELED]. Its ring-closure residual is about 2 × 10⁻¹⁵ kPa, which is floating-point noise: it confirms the supplied network was recomputed consistently and confirms nothing about whether that network represents the plant.
Limits and open items
Modelled and not frozen: the 1.177, the 275 kW residual, the 14,125 kW facility power, the 1,686 kW overshoot and its attribution, the 875 / 642 kW pair, and the 609 kW earlier record. [MODELED]
Open: the four interfaces that make 88.4 % of the overshoot; the weather recomputation, which this revision did not perform; the measurement window and weather band the contract ratio is scored against; and an independent solver for the network check. [HOLD]
No meter produced any figure here. No generator was sized from this table. Nothing in it establishes that the contract basis will be met or missed. Any published ratio still has to name its meter point, its weather band, its liquid fraction and OEM branch, and that the weather chain is a placeholder with no site validity. This is a robustness argument about a number still pending recomputation.
© 2026 K&K Data Service Inc. All rights reserved. Reproduction or republication is permitted only with clear attribution to K&K Data Service Inc. and a working hyperlink to the canonical URL of this article. Excerpts must preserve the technical context, maturity labels, assumptions, and limitations. No excerpt may imply project approval, field validation, certification, or endorsement that the original article does not state.
Source: K&K Data Service Inc., “A Modeled PUE of 1.177 Is Not Robust With a 275 kW Residual,” https://www.kkdatasvc.com/lab/heat-rejection-and-pue/a-modeled-1-177-is-not-robust-when-residual-is-275-kw/.
Request the Public Calculation Note. If your project record shows a different result, or the same failure domain under another name, we want to see it. Email inquiry@kkdatasvc.com or use the contact page.