Dry-type transformer failures versus liquid transformer failures
In this "Cut the Copper" installment, Guentert explains how, when properly installed, liquid-filled transformers in data center power systems are more reliable than air-cooled dry-type transformers.
I stated at the outset of this series that it would be controversial, and might receive a backlash of unfavorable response. Here is my most controversial statement, as clearly as I can state it:
In my opinion, after having carefully watched this situation develop over 35 years of my professional life, I’ve concluded that properly designed liquid-filled transformers, when installed in properly designed data center power systems, are more reliable than air-cooled dry-type transformers, when switched fairly frequently by upstream vacuum breakers, at voltages of 10 kV and above, by a factor of at least 50.
How’s THAT for being politically incorrect? Whenever I make this statement to a new client or to a group I’m presenting to, the immediate response is usually, “Show us some hard data to back up that terribly assertive claim you just made.”
The hard data don’t exist, I admit. I think that this is mostly because no data center owner ever reports its failures to any sort of a registry or information clearing-house, nor do transformer manufacturers want to report their failures, either. All I can do here is report my own personal experiences:
--I have personal knowledge of more than 75 failures of dry type transformers installed in data centers (95% of them VPI’s) under the switching circumstances described previously, over 35 years of experience. I’d have to believe that there are many, many more I haven’t heard about.
--I’ve personally investigated more than 30 failures of dry-type transformers under these conditions, at various sites—some at the request of my clients and employers, and some just out of my own curiosity.
--I have personally witnessed the failures of 5 indoor dry-type substation transformers at the instant of occurrence, while standing with 30 ft of each of them at the time, inside electrical rooms of active data centers. (Blinding flash, loud “BLAM!!!,” small flames, heavy black smoke, fire alarm callouts, fire companies running into electrical rooms with water hoses. Two of these failures occurred simultaneously, with a single breaker closure, in one data center.)
--I work closely with a select group of large general contractors, who specialize in construction of large data centers, and who have probably collectively built at least $8 billion of large new data centers in the past 5 years. Their guys know of my passion for this topic, and I generally receive about 5 or 6 phone calls or e-mails per year from them. E-mails like, “Hey, Joe. Two more today, at a data center in Virginia, about 1:30 p.m. Simultaneous loud flashover of two 2.5 mVA 12.47 kV VPI units, with a single breaker closure, during final Integrated Systems Testing. No snubbers were installed.”
--I’m aware of at least two manufacturers of dry-type and cast coil transformers who declared bankruptcy and closed up shop, under the weight of warranty claims and litigation from similar failures.
I’ve observed a fairly recent shift in attitude of the remaining major dry-type transformer manufacturers. On a failure of a 2.5 MVA VPI unit in a data center just two months ago, I heard the manufacturer’s response to a request to IMMEDIATELY provide a replacement transformer at no charge under warranty terms. Paraphrasing slightly:
“Warranty claim replacement is denied. This is not a manufacturing defect in our transformer covered by warranty…it’s a widely known systems problem with dry-type transformers switched by upstream vacuum breakers. There was no snubber installed on this transformer. Go speak with your consulting engineer about his system design problem.”