Fifty kilowatts of air-side terminal fan load — computer-room air handlers, rear doors and air walls — was written into two step registers of the same table. The controlled erratum records the defect against the generator staged-capacity table: the load sits in the uninterruptible-supply step and again in the first mechanical step, so one physical load is counted as 2 × 50 = 100 kW and the staging total is overstated by 50 kW [MODELED]. The physical fan load never changed. What changed is what the table said about it.
Each Load Item Must Belong to Exactly One Step
The risk paragraph gives two consequences and grades the second as the more concealed. Inflated capacity drives oversized equipment selection and inflated investment, and makes the margin check return a false verdict of ample margin. Less visibly, the controls basis cites this table directly for its load-staging sequence; if the staging logic in the power-monitoring and building-management systems is set from the inflated value, the staging criteria in service will not match the real load. No such logic has been set in the read set — this is a citation path, not an observed misconfiguration.
Over- and under-statements coexisting in one load table is not a question of how conservative to be. It shows the table lacked a structural constraint: each item belongs to exactly one step. [FACT]
After Correction, Generator Transient Is a Rectifier Problem
The erratum declined to pick the home. It reserved that to the control editor under the existing step definitions and required only that one step be chosen and both not coexist. The heat-rejection revision then made the assignment: unique home is the uninterruptible-supply step, because the item is already inside protected load and therefore already inside rectifier input demand.
After correction the staged table reads 12,968 kW for the uninterruptible-supply step, which already contains the 50 kW; increments of 220 kW, 275 kW and 405 kW for the three mechanical steps, totalling 900 kW against 950 before [MODELED]. The first mechanical increment had been 270 kW. The cumulative figure of about 14,125 kW reconciles with the baseline-column facility power in the load chapter — a reconciliation between two internal tables, not an external verification, and the generator counts behind the percentages remain research figures with the branch unadjudicated. The 82.3 % share is against 15,750 kW of generator N+1 available capacity [MODELED], itself a research figure at an unfrozen site derate.
The load-table author had used this table to make a negotiating point, and the correction strengthened it. The uninterruptible-supply step alone is 82.3 % of that N+1 available capacity, while all mechanical load across three steps is 5.7 %. The stated consequence is that this project's generator transient problem is almost entirely a rectifier problem and not a cooling-equipment problem — reversing an expectation set earlier in the same document, where the concern had been that mechanical loads staged last would lengthen the loss-of-cooling window.
The Double Count Had Overstated Early-Staging Cost
The cooling side's capacity cost of being staged early was quoted as 2.1 % of the rectifier load before the correction and 1.7 % after it [MODELED]. The author flags the direction explicitly: the correction is downward, meaning the earlier revision had overstated that cost, and the conclusion's direction and strength are not reversed but strengthened.
That is what the double count was hiding. It was not hiding spare megawatts; it was hiding how cheap the cooling side's request already was — a reading of the stated direction rather than something either source says in those words. Putting 220 kW ahead of the 12,968 kW walk-in is almost free for the generator and expensive for ride-through if delayed. The associated change candidate is upgraded in the same pass from a joint decision item to a joint decision item with a recommended value: stage the first mechanical step ahead of the rectifier walk-in, pending an outstanding Engineer of Record verification that has not been performed.
Three Table Rules Forbid a Load Sitting in Two Steps
Three table-level constraints were added rather than a single value correction:
- Each load item's step assignment is single-valued, and wordings of the form "prefer step A, failing that step B" are prohibited; where recovery order needs expressing it belongs inside one step as a priority column.
- The uninterruptible-supply step is carried as the whole rectifier input demand, so any item already counted into the protected load may not reappear as an increment in a later step.
- The table total must equal the sum of items under unique assignment, and any item appearing in two sequences is a check failure.
Step assignment is not recovery priority. The air-side terminal recovers more slowly than the liquid side because its thermal inertia is larger, and that lower priority is what tempted the dual assignment — an attribution of motive the source does not make.
The Three Errata May Not Be Netted Against Each Other
Three numerical errata in this batch move in inconsistent directions: a remainder-correction, a charge-channel fix of plus 46.8 kW [MODELED], and this minus-50 kW unique-home correction. They may not be netted against one another, and must be published together or an intermediate revision will carry an internally inconsistent load table. [FACT]
If the reserved adjudication lands on the first mechanical step instead, the 220 kW increment moves again and the staging recommendation must be rechecked against it.
Limits and open items
Confirmed on the current record: one physical 50 kW load was written into two step registers; unique home is the uninterruptible-supply step; the three numerical errata may not be netted. [FACT]
Modelled and not frozen: the 12,968 kW uninterruptible-supply step, the 220 / 275 / 405 kW mechanical increments, the 900 kW against 950 kW totals, the 82.3 % and 5.7 % shares, the 2.1 to 1.7 % revision, and the 15,750 kW N+1 denominator. [MODELED]
Open: the Engineer of Record verification behind staging the first mechanical step ahead of the rectifier walk-in; the generator machine count and nameplate; and the reserved step assignment if a later editor reverses the unique-home ruling. [HOLD]
No generator quantity is released. No staging sequence is sealed. Who signs the assignment?
© 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., “50 kW of Fans Was Counted Twice in the Generator Step,” https://www.kkdatasvc.com/lab/heat-rejection-and-pue/50-kw-of-fans-was-counted-twice-in-the-generator-step/.
Discuss This Constraint. 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.