SAP Data Center to Go DC to Save Big Bucks

December 12th, 2010 by Susan Kraemer

Michael Kanellos at Greentech media has a story about an interesting concept for energy reduction that is new to me: switching a data center to DC power.

SAP is making some energy retrofits to save money, and while the three biggest savers are the usual ones that we all know about (substitute a solar array for utility power, switch lighting to LEDs, cut airfares by videoconferencing, ) the fourth caught my attention. SAP is converting their data center to DC.

Solar panels make power in DC current, and they lose some efficiency by having to be converted (that’s what inverters do) to AC. So this will leverage the investment they are also making in solar: about $1.2 million. Like most businesses who get their own renewable power, they will save big time just by making their own energy with solar.

But, their planned solar array would produce about 15% more if it could be used without an inverter by being delivered directly in its native DC current. A 24 KW DC system will deliver about 20 KW of power once converted to AC. Conversion to AC loses about 15%.

They figure that with this move to go DC in the data center, if the solar was able to deliver its DC current straight to the data center without any losses due to conversion to AC, that would save as much as 40 percent of their power consumption.

Without their own solar, but just regular utility AC, they calculate conversion to DC for the data center would save only 15 to 20 percent.

This is an exciting experiment. Nobody has tried this, to my knowledge, at least not on this scale. Off-gridders and RV-ers use DC appliances so that one tiny solar panel can more efficiently power a computer and a fridge etc, but SAP is on a different scale.

Alternatively, just to convert the data center to DC current, without going solar, would involve installing a ” rectifier” that can convert grid AC power to DC to run the computers and storage equipment. DC rectifiers save power by reducing the number of times power gets converted from AC to DC and vice versa before it powers a server.

But if they could eliminate the step of converting to AC, using solar off the roof, the DC rectifier could take a nap and just take raw solar power in its natural state: DC. This would save money twice: no $128,000 rectifiers to convert grid AC to DC, plus no inverters to convert the roof solar to AC, before it gets “rectified” again back into DC.

Worth taking a look at how this could be done: an exciting experiment.

Image: Daniel Howherd

Susan Kraemer@Twitter

Keep up to date with all the hottest cleantech news by subscribing to our (free) cleantech newsletter, or keep an eye on sector-specific news by getting our (also free) solar energy newsletter, electric vehicle newsletter, or wind energy newsletter.

Tags: data centers, DC, SAP

About the Author

Susan Kraemer writes at CleanTechnica, CSP-Today, PV-Insider , SmartGridUpdate, and GreenProphet. She has also been published at Ecoseed, NRDC OnEarth, MatterNetwork, Celsius, EnergyNow, and Scientific American. As a former serial entrepreneur in product design, Susan brings an innovator's perspective on inventing a carbon-constrained civilization: If necessity is the mother of invention, solving climate change is the mother of all necessities! As a lover of history and sci-fi, she enjoys chronicling the strange future we are creating in these interesting times. Follow Susan on Twitter @dotcommodity.

Bob Wallace

Something isn’t sounding right here. From what I know DC to AC inverters are more like 95% efficient.

http://www.affordable-solar.com/xantrex-gt-50-inverter.grid.intertie.htm

http://partsonsale.com/Soleil_2000-120_2000-240_SlickSheet%5B2%5D.pdf

SAP can save themselves money by installing enough solar panels to provide 100% of their power, but that system will end up selling most (~80%) of its produced power to the grid (assuming a 5 hour solar day). Then SAP will be able to buy get the power they need for the times when the sun isn’t shining.

During the hours that the sun is out they could feed their equipment directly from the panels and avoid a conversion loss (5% to AC, 15% back to DC).

But they are going to be sending many more times as much power to the grid and that must be converted to AC. And then when they get that power back from the grid it’s going to be coming back as AC – so a conversion loss for 75% of the day that can’t be avoided.

I suspect a better way to describe what they are doing is to say that in order to produce 100% of the power they need from their own solar panels they would have to install ~120% as much wattage capacity as they use. The producer/user is the one paying for conversion loss. They are going to loose ~5% going from panel to grid and ~15% going from grid to operational DC.

By not sending a portion of their panel DC to their internal grid they can avoid that ~20% loss. For the 25% of the day in which their panels are making power. Sounds to me like their plan will save them about 5%, not 20%.
- http://cleantechnica.com/author/susan Susan Kraemer
  
  Hi Bob, re loss of 5% versus 15%, they vary: Sunpower panels lose very little in the inverters, likely in your range.
  
  But I was going on the California Solar Initiative ratings, where you can check individual panel manufacture and see what AC winds up being – per the CEC (California Energy Commission) ratings. I checked Sharp, as middle of the road.
  
  Maybe the CEC errs on the side of caution, after all they pay us the rebate based on the expected performance, showing what your panels will likely contribute to the grid in AC. http://www.csi-epbb.com/
  - Bob Wallace
    
    This page?
    
    http://www.gosolarcalifornia.org/equipment/inverters.php
    
    The first three entries are high wattage inverters with 97.5% efficiency ratings. A quick scan seems to say that almost all the listed inverters are 95% or better. Why would anyone go with a lower efficiency model and waste that power?
    - http://cleantechnica.com/author/susan Susan Kraemer
      
      No, sorry, I used to do solar estimates, and the only page I ever used was their calculator: it’s this page
      http://www.csi-epbb.com/default.aspx – should say CSI EPBB Calculator you can enter the make and model of any solar module there and find what the CEC rates any of them in AC, compared to their DC rating.
http://electric-vehicles-cars-bikes.blogspot.com/ Paul

40% is HUGE!

Common PC power supplies are only 70 75% energy efficient so removing the AC transformer for 1,000s of servers would not only improve their energy efficiency but would also remove a significant heat source which results in less energy being required for cooling systems.
Joe

Look, if you want some credibility, you need to have articles written by people who understand their topics. This author knows nothing about this issue, and the result is that there are a lot of words on the screen, but none of my questions was answered.

She appears to believe that you can run a wire from any DC source to any machine and . . . Bob’s your uncle!

Will this facility have several very large low voltage (large gauge) buses serving individual computers that have no usual power supply? Will the computers be built to use a single voltage for all components? How will individual machines be isolated from others or will there be novel protection schemes?

It requires more than, “a rectifier,” to convert AC to DC.

This article belongs in a high school newspaper.
- Shounak
  
  I have read up a lot about DC datacenters. Invariably articles that I read are optimistic but every one of those articles have comments below that rubisshes the claims. Whats the truth ? Why is there so much ambiguity on this one ?
  - http://cleantechnica.com/author/susan Susan Kraemer
    
    @Shounak, that is not anything unusual. All renewable energy topics seem to incite comments that “rubbish the claims.” There is a Luddite mentality out there, though one wonders how or why they frequent cleantech sites. Inventors, thinkers and product developers are apparently a very tiny portion of the population. The mindset of being open to new ways of doing things is seemingly quite rare.
Michael

Please don’t post. I just read the original article and understand they are not eliminating the rectifiers and will use the grid for stability and night operations.
Michael

This sounds great. How are they storing power for night operations?
- http://cleantechnica.com/author/susan Susan Kraemer
  
  I am sure they will wind up still sending some at least for storage on the grid to be able to credited some power for night operations, but my reading of it – the exciting and novel part – is that they are investigating using at least some of the solar straight, without converting it to AC – but rather, making the servers able to use the solar in DC.
  
  What I was unable to find out was what percentage of their total power is to come from their solar. It will be an interesting story to follow. If it can be made to work on this scale it would be huge. How much the equipment needed costs, how much of the solar can be used straight, versus sending it to the grid as usual, etc, lots of questions.
  
  A very interesting experiment.