The early 2000s: The amazing boom of data center construction
Data center engineers have split into two camps. Which camp are you in?
Prior to 2000, almost any data center in the world that had 10 mVA of total load would have been considered a very large data center. Mostly due to the growth of the Internet and cloud computing over the next 10 years, data center total loads grew from 10 to 20 to 40 to 60 to 80 mVA, and some data centers are being planned now that will have more than 100 mVA of total aggregate load.
This dramatic growth presents great challenges for design engineers to move that much infrastructure power into a data center economically, efficiently, and safely.
The loads have grown so much and so fast that many utilities can no longer serve these loads with their present medium-voltage distribution systems, and more than a few owners are now having to take utility service at transmission-level voltage, and construct their own primary substations. The loads are now often so large that it has become impractical in many cases to use even 15 kV class distribution systems–a distribution voltage of 15 kV is just too low, in many cases. Feeder cable sizing and breaker ampacity ratings often become too large at 15 kV, and short-circuit duties can become excessive for downstream equipment. Sometimes, the only good reason to continue to “force” distribution into an ever-shrinking 15 kV “box” is to avoid another transformation for diesel generators, so that utility-generator transfer operations can be made directly at the 15 kV level.
As this trend has developed, it appears to me that the consulting engineering community has more or less split into two camps:
- One camp prefers to stay with highly reliable liquid utility transformations, located outdoors, and then run large underground duct banks into secondary switchgear located inside the building.
- The other camp remembered the lessons learned from the “World War II loadcenter” concept, and moved their transformations inside the building, physically closer to the actual loads.
Here’s an at-a-glance economic comparison between the two different approaches:
Figure #3 – Comparison of Approaches
The example shown is not very typical, because in the 480 V example, most engineers would do all possible to place the transformers as close as possible to the secondary switchgear, in order to minimize the lengths of the 480 V conductors. But, the example is nonetheless real-life, and I’ve seen this very arrangement in a few projects.