報告_OCP2026_GammaTech_2.3MW_CDU設計比較_20261015

PDF 原檔：報告_OCP2026_GammaTech_2.3MW_CDU設計比較_20261015_original.pdf

講者公開簡報（OCP 官方 Recorded Content 頁提供下載）。文字以 PyMuPDF 逐頁抽取，### p<n> 為頁碼；圖表內容見原檔。

圖片清單（已驗證 2026-10-19）

原始內容

p1

Jake How Application Engineer, Gamma Technologies Tommaso Ferrarese Business Development Manager Data Center Cooling, Alfa Laval Holly Hilten Account Manager, Data Center, Alfa Laval In collaboration with Stulz A Comparative Study of Design Choices for 2.3 MW Systems What Shapes CDU Performance? COOLING ENVIRONMENTS

p2

One capacity target. Four performance goals. THE QUESTION 2.3 MW held fixed in every case 1 Lowest capital cost What is the least expensive hardware that still delivers the duty? 2 Lowest electrical consumption What does it cost to run, across a whole year? 3 Effect on pipe sizing How much primary flow does the selection force? 4 Effect on pump sizing What head and flow must the facility pump deliver? The question. How do we select equipment to optimise these four outcomes, and build a tool that reproduces the process when the external variables change?

p3

Different Types of Heat Exchangers WHAT WE LOOKED AT Brazed • Compact and light • Easy to install • Self-cleaning, little maintenance Gasketed • More efficient, lower OPEX • You can add plates later • Openable, serviceable, re-gasketable Sizing Basis • ASHRAE/Deschutes • 1-4 HX • 1-2 Passes • 1°C-4°C Approach • CB210, CB450

p4

CAPEX Comparison QUESTION ONE

p5

A CDU model calibrated to Deschutes 5 METHOD, 1 of 3 HX FCV Exp. tank Pumps Built and calibrated against the Deschutes 5 specification so that every alternative is compared to a known reference rather than to an assumption. Parameter Reference value Capacity 2.0 to 2.3 MW Approach (ATD) 3 K TCS flow 500 GPM FWS flow 500 GPM TCS fluid PG25 or treated water FWS fluid Water Heat exchangers 3 brazed, two-pass, parallel Pumps N+1, sealless *results are per HX

p6

Common Evaluation Criteria for Comparison METHOD, 2 OF 3 ANSI/ASHRAE Addendum b to Standard 127-2020 approved by ASHRAE and ANSI on 27 February 2026, bringing liquid-to-liquid CDUs under a method of test for the first time Parameter Standard value Fluids Water on FWS, PG25 on TCS FWS supply 26 °C TCS supply / return 30 °C / 39.6 °C at full capacity TCS flow Held at rated across all load points Part load 75 / 50 / 25 % by TCS entering temperature Fouling As-tested fouling factor is zero Location Test cell only, not field Why we switched to it • Holding the ASHRAE conditions fixed instead makes every exchanger comparison apples to apples. • The secondary side is then fully constrained, so the whole effect of the selection appears on the primary side. That is what the next slides measure. Conditions paraphrased from the addendum. ASHRAE distributes addenda and errata separately, so verify currency before the final submission.

p7

The CDU does not act alone, so we modelled the plant METHOD, 3 OF 3 A standalone CDU tells you very little With the secondary side fixed by the ASHRAE conditions, the selection cannot change the IT-side result. It can only change what it demands of the plant. So, the exchanger was integrated into a full system model and everything outside the CDU was held constant. What an integrated approach achieves One apples-to-apples question, asked eleven times: for the same duty at the same rating conditions, what does each exchanger demand of the primary loop, and what does that cost to run across a year? What the system model contains • Primary side: dry cooler, air-cooled chiller, buffer tank and facility pump • Secondary side: fixed IT load of 2.3 MW • Secondary pressure drop fixed at 50 psi • Estimated dry cooler approach, CoP • Estimated chiller CoP as function of load, air temp, liquid temp • Estimated FWS pump map

p8

Baseline Comparison: Results @25C Ambient Deschutes Baseline Brazed ASHRAE, 2HX, 1 pass, 4K Gasketed ASHRAE, 1HX, 1K Deschutes Baseline Brazed, Deschutes, 1HX, 2 Pass, 3K Brazed,Deschutes, 2HX, 1 Pass 4K Brazed ASHRAE, 2HX, 1 pass, 4K Gasketed ASHRAE, 1HX, 1K Gasketed ASHRAE, 1HX, 1K Brazed ASHRAE, 2HX, 1 pass, 4K

p9

Evaluation of each power consumer as a function of ambient temperature ANNUAL BEHAVIOUR At low ambient every selection looks much the same, because the dry cooler is doing all the work and pump power is a small share of the total. Above the changeover the chiller takes over, and the selections that return the primary water hotter keep the compressor off for longer. Brazed ASHRAE, 2HX, 1 pass, 4K Gasketed ASHRAE, 1HX, 1K

p10

Annual Energy Consumption for Different HX Selections ANNUALISED ENERGY Temp Bin (°C) Hours -12 to -10 2 -10 to -8 16 -8 to -6 18 -6 to -4 47 -4 to -2 74 -2 to 0 117 0 to 2 161 2 to 4 205 4 to 6 221 6 to 8 351 8 to 10 247 10 to 12 396 12 to 14 426 14 to 16 395 16 to 18 594 18 to 20 509 20 to 22 782 22 to 24 947 24 to 26 788 26 to 28 859 28 to 30 469 30 to 32 542 32 to 34 424 34 to 36 131 36 to 38 37 38 to 40 2 TOTAL 8760 Dallas, TX Results will vary by location: • Ambient temperature • Price of electricity (impact on OPEX) Other site-specific considerations also effect results: • Actual load vs. nominal capacity • Evaporative cooling • Performance of external equipment Deschutes Baseline Brazed ASHRAE, 2HX, 1 pass, 4K Gasketed ASHRAE, 1HX, 1K

p11

Three variables the rating point does not cover OFF-DESIGN Tighter approach 3 K instead of 4 K Primary-side fouling 0.9e-4 m²K/W Different primary fluid PG35 instead of water Relative Increase in Energy Consumption HX 3K Approach Fouling PG35 Deschutes Baseline -1.7% 3.0% 2.6% Brazed ASHRAE 2 HX, 1 pass, 4K -0.5% 7.0% N/A Gasketed, ASHRAE, 1HX, 1K -2.1% 0.4% 0.4% cannot supply target temperatures

p12

An easy way to run this analysis REPRODUCIBILITY

p13

What we would take away from this WRAPPING UP 1 2.3 MW can mean very different machines Every one of these ten does the same duty. Change the operating conditions and you get a different box: return temperature ran from 34 to 39 °C and primary flow from 660 to 940 GPM. 2 That changes your pump, your pipe size, your CAPEX and your cost to run. Capital cost spans a factor of three. Annual energy spans about 17 %. Flow crosses two pipe classes and drags the facility pump with it. The cheapest one to buy is not the cheapest one to run. 3 So design and optimize the cooling loop as a system. Once the conditions are fixed, the heat exchanger cannot change what the IT equipment sees. All it changes is what it asks of the plant around it. Judge it there, over a year, and not on a single number on a datasheet.

p14

• Off-design belongs in our specifications. • Specifications typically list a CDU at one design point. This study demonstrates interesting behavior in other conditions. • Let us agree how to evaluate performance. • A shared way of doing ambient sweeps, fouling allowance, alternative fluids and part load would let operators compare CDUs on what they actually cost to run. • Publish the operating point, not just the megawatts. • Many other factors beyond the capacity are required to properly characterize the CDU performance. • Let’s discuss! • Please continue the conversation at the summit or in the CDU workstream. Call to Action

p15

Thank You Jake How Application Engineer, Gamma Technologies J.How@gtisoft.com Heat exchanger selections and technical review contributed by Stulz and Alfa Laval.