Slashdot Mirror
More Data Centers Using On-Site Solar Power
1sockchuck writes "Solar power hasn't been widely used in data centers because it takes a very large installation of photovoltaic solar panels to generate the levels of energy required by these facilities. But the month of April has seen the debut of four new data centers featuring on-site solar arrays."
While I don't think it would be feasible to run a data center only on solar, it could help with a big thing: cooling. The hotter it is outside, the harder those A/Cs have to work and the more energy they use. Well, conveniently the hotter it is outside the more direct sun the solar panels tend to get so the more power they generate. Kinda of an automatic offset. When the power demand is the most, the panels give you the most.
You'd still need line power to run the data center, particularly at night, but you could help offset your costs in a big way.
They dont need to last for 12 hours, just long enough to make the transition to grid power.
It's not necessary for solar to cover the entire power needs of a data center. It'd be nice if it did, but any power generated is money saved on their electric bill (and less drain on the general grid). And as stated above- they already have lots of batteries to cover (if) any surplus generated, and a fairly constant demand.
Anyway, one can only hope the trend will continue, even if only for two very selfish reasons:
a. the more mainstream the PV are, the lower the price on all the market (10 years to ROI for a decent PV home installation is still too expensive to my taste).
b. the more pressure on energy consumption to run a data center, the higher chances computer (part) manufacturers to research techs with lower energy requirements.
I reckon both of them would be good (medium/long term) for my pocket as well.
Questions raise, answers kill. Raise questions to stay alive.
This is a roundabout way of going about it unless you power an entire datacentre off DC. Solar is typically stepped up via an inverter for larger operations, and the inverter syncs with the grid. Its actually kind of fun to watch our old analogue power meter stop spinning backwards and start spinning forwards when we turn the oven on and stop exporting power.
The transition between night and day should be perfectly seemless with no batteries required. These aren't diesel generators that we need to wait for, it's a grid, there's no sync delay and every Elec101 student should be able to design a system that seemlessly goes from export to import with no need for batteries.
That's not to say that there don't need to be batteries, but that switching to grid power is definitely not one of the reasons for them.
This hits the nail on the head. Solar PV to offset during the day and grid power at night. I'm sick of Greenies pretending that Solar is the panacea of all things power related, and sick of the short sighted people saying it has no place in power and that's the end of story.
This is where we should be heading. More solar power in plants with a large HVAC requirement. So during the day when it's hot you can offset the cooling energy required, and the result is reducing the grid power which is attainable, rather than replacing base load generation with some magical fairy green power station... which is not!
Solar power is a perfect match for data centers. Their power demand is basically constant
Especially if you built it on an asteroid with no clouds and with a side always facing the Sun. Cause you don't need those coal and nuclear plants generating base load, and those huge dams regulating peak load, like on Earth.
And how would you go about making HVAC more efficient? There's no magic to it, companies keep working on better technologies but it is what it is and there's no magic method to better efficiency that we could use but don't. I would bet data centers invest in efficient HVAC equipment since that is a significant cost savings but there's just only so good you can get. For large facilities about 14 EER is as good as you can get for your equipment. If you want it higher you have to water cool the A/C's condenser coil. That can be done, but of course requires a ready supply of water for the chillers. Over all effective efficiency can be improved with variable capacity units, either having more smaller compressors that can go on and off or using variable digital scroll compressors like AAON uses but that's it.
There are just real limits to A/C efficiency and there is plenty of research going in to trying to make them better, but it is incremental improvements and whatever the best technology is, companies bring it to market.
And how would you go about making HVAC more efficient?
How about, instead of making electricity at 10-15% efficiency and use it to run a 20-30% efficient refrigerator, switch the whole datacenter to absorption cooling which uses solar heat directly ?
This way, instead of an abysmal solar efficiently of 5% and large capital costs, you get 20% efficiency with some piping and thermal captors.
There is solar airconditioning at industrial scales. Refridgerated AC is a heat pump after all and solar heat is about as good as any other once you get away from the small scale domestic units.
This must only be bullshit. Everybody knows solar power is a pipe dream and will never be viable. Data centres should be burning good ole coal and oil, or even tires.
This solar power silliness reminds me of those crazy dudes in the past wasting all taxpayers' money to invent flying machines. Or cures for infections. Or transmitting images over the air in "invisible" waves. All pipe dreams.
There's a reason for this. In large systems the need for a fast switch from utility to inverter power under full load requires that there be a battery charger constantly charging a battery bank and an inverter or rack of inverters connected to the battery bank to provide power to the load. This isn't very efficient but it works well.
Also battery voltage can vary from about 10 V to 14.4V per battery so I don't think this is tightly regulated enough for computer components. In a large system you want to keep the DC voltage high so your I^2*R power loss is minimalized so DC voltages as high as 108 V are not unheard of.
Mod me down with all of your hatred, and your journey towards the dark side will be complete!
Maybe not Elec 101 but 401 (at least in the US). The idea isn't that complicated but there are regulations that make putting a system like that into practice a bit of a challenge (at least in the US). The system must go off-line if the phase drifts, if the line voltage changes by x in y milliseconds, etc...
Mod me down with all of your hatred, and your journey towards the dark side will be complete!
Whoooosh is the sound of a flying machine going over your head. This machine is made of a metal called Irony.
That over the next 20 years, there will be no advances in technology leading to better efficiency and lower cost.
Every tech when starts getting adopted is expensive. For example, could anyone afford a computer at home in 1960s?
As technology advances, and costs come down, solar will become more and more viable. So to dismiss a tech just because currently its not economically viable is basically foolhardy.
My Aurora : http://www.youtube.com/watch?v=o91ZsGwJYyg
FB : https://www.facebook.com/TanveersPhotography
Because absorption coolers don't perform well, especially for heavy loads?
The first one was done without taxes. The real developments that made them useful commercial tools were mostly done in support of....get this....military operations. That was the federal government doing what it was supposed to be doing.
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
Here's the skinny -
PV Costs about $5 per watt, installed (and that's cheap. It's usually much more). Never mind the cost of the silicon technology, the majority of that cost is in brackets, adhesives, other hardware, wires, labor... Costs which are fully mature and not going to go down. So, assume an average upper middle-class house in south Texas that uses roughly 100 kWh per day (Central A/C in a large-ish house is a bitch!) and forget about batteries. So a 10 kW system is going to cost us $50,000 (10,000 watts * $5) and generate 60 kWh per day, during the daytime only (6 hours of good direct sun). We still need grid power to get to our 100 kWh usage.
Similarly, that Grid power only costs $0.12/kWh, so the same power from the grid only costs us $7.2 per day vs. the $50,000 solar array. (And trust me, when it's 105 degrees in the summer time, for months at a time, that $7.20 feels damn cheap.)
Going further, $216.0/month for grid power, vs. a $330/month payment on a 20 year loan @ 5% interest.
Or, better yet, invest your $50,000 that you already have in an index fund, get an average 7% return and EARN $390 per month over the next 20 years. Now your grid power is free and you make an extra $174/month, and have zero maintenance worries (dust, corrosion, etc.) or insurance costs (hail, baseballs, etc.)
And that's only in the summer time. My winter electricity usage is about half that, which makes the solar picture even bleaker.
People who value their money more than their conscience will never buy solar, until the cost of grid power essentially doubles. And to be truly competitive, the cost of grid power needs to triple. But what would that do to our economy when everything you buy has power as one of its major embedded costs?
My power has tiered pricing where the more I use the more it costs per kw*h. I've done the math and for a few thousand bucks I could put together a solar array that offsets that *more expensive* power and could pay for itself in a hand full of years. Generating enough power to cover the lower priced energy is not cost efficient though. YMMV
Check out my lame java blog at www.javachopshop.com
Yes, but unless the panels are attached directly to the building, there will be a cooling air gap to keep the heat away from the building. Besides, most roofs are already black.
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
I support the solar power *movement* but this simply isn't true. It costs about $30k for me to replace my current electricity usage. At an average power bill of $150/mo it'll take 200 months or 16.7 years to pay off. That assumes I have $30k lying around I don't need. Now if energy costs triple in 5 years I'll be quite sad that I didn't get that solar system, but even putting that money in a pitiful money market account at the local credit union will offer better returns than the PV system today. And the net carbon usage to mine and manufacture the solar panels still isn't that good.
Check out my lame java blog at www.javachopshop.com
Does anyone see this as anything other that a PR stunt? Facebook's datacenter uses 30MW of electricity -- a 100KW solar panel array will produce 0.1% of their power - not even a drop in the bucket. (note that it's not 0.3% since the solar panels don't provide power all day).
If they were really interested in reducing their carbon footprint with solar, they'd be investing in one of the large-scale power plants being built in the desert where they can buy more KW per dollar. it doesn't matter whether they reduce carbon in Arizona or in Oregon, it's all the same to the environment.
And if they were *really* interested in reducing their carbon footprint, they'd use a small nuclear reactor to generate 100% of their power on-site. Which would make a *real* difference in their carbon footprint rather than a meaningless symbolic gesture.
That is wrong, likely it is a typo.
Regards
angel'o'sphere
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
Google powers its servers off 12V DC with inbuilt batteries on each server. The savings in power costs make this worth it. http://insidehpc.com/2009/04/02/google-unveils-its-super-secret-server-design-dc-and-batteries-built-in/
They don't perform well? Isn't that just a question of sizing? If they don't work well for heavy loads, why is it that there's lots of them available in huge sizes, but hardly any residential sized units?
Besides, consider the context, even if adsorption chillers aren't as efficient as HVAC versions.
1. Solar heat panels are cheaper than solar electric panels.
2. Solar heat panels are more efficient than solar electric (30-90%, vs 13%)
Basically, the cheaper panels helps to offset any increased costs on part of the adsorption system, while the higher efficiency helps offset any losses in efficiency.
I don't read AC A human right
There's a reason for this. In large systems the need for a fast switch from utility to inverter power under full load requires that there be a battery charger constantly charging a battery bank and an inverter or rack of inverters connected to the battery bank to provide power to the load. This isn't very efficient but it works well.
I think this is partly what he was talking about. IE, instead of having a UPS which converts battery power to AC while simultaneously charging the batteries, just have a big DC power supply, float the batteries on it, and use that to power all of the servers and stuff. This does make sense; I have a similar setup with a Mini-ITX motherboard I use as a home mail server and some other things - it has the 12 volt car power supply in it, along with a small 12 volt gel cell battery for backup.
Also battery voltage can vary from about 10 V to 14.4V per battery so I don't think this is tightly regulated enough for computer components. In a large system you want to keep the DC voltage high so your I^2*R power loss is minimalized so DC voltages as high as 108 V are not unheard of.
I think there are commercial systems that operate with DC like this; I know for the telecom 48 VDC is used a lot. This is still a manageable voltage, as high voltage DC is a little more difficult to work with than AC at similar RMS voltages (though I'm sure higher voltages like 108 VDC are used too), but you don't need absolutely huge wire. Also, there are low voltage DC-input ATX power supplies available. I've seen them for 12, 24, and 48 volts. Generally they will produce a regulated output for the PC, so if the battery voltage fluctuates somewhat it's not much of a problem, provided it's within spec. Also, normal AC units usually work by rectifying the incoming line to DC, and then using a step-down converter to get the lower DC voltages. So you could probably run one directly off of 170 VDC (170 volts is around the peak of the AC sine wave).
And of course, if you have a system like this it wouldn't be that hard to throw some photovoltaics in too, which would reduce the load on the main power supply (and thus the draw from the grid).
Saying "I'll probably get modded down for this" in a post is the best way to get it modded up.
Cheap solar thermal panels do not run at high enough temperatures to make absorption refrigeration efficient.
Plus, the adsorption chillers are, comparatively, solid state. Little to zero cost to run or maintain, so all capital costs. Sounds like a no brainer - to at least handle the 'base load' of cooling (so to speak), and run the HVACs as needed only.
cheaper != cheap.
I never said that they'd be able to use bargain basement thermal panels.
Still, Evergreen Es-A-210-Fa3
210 watt, 13% efficiency, 65"x37.5" = 2,438 square inches of collector(1.57m2). $587 each bought in a pallet of 28.
Cost by area: $374 m2
Power by area: 134 w/m2
Cost per watt: $2.79
A "ThermoPower-VDF30 should be around $1,244.
It's aperture is 2.67 m2. 94% absorbtion, 7% emission per the datasheet. 'Stagnation Temperature" > 428F
Cost by area: $466 m2 (okay, it's not cheaper per m2)
Power by area: 634 w/m2 (at 1000 w/m2) (almost 5 times the power though)
Cost per watt: $0.74
Conclusion: At 27% of the price, you'd have to be looking at a pretty hefty penalty for using an adsorption chiller for it to make sense to use electrical HVAC with solar electric panels. Also, in either case I'd consider utilizing additional reflectors to put more power to the relatively expensive collector.
I don't read AC A human right
Don't even get me started on the space program. What a waste of taxpayer dollars! In less than 50 years after we wasted billions to put a man on the moon, the much superior PRIVATE sector has already put a person into low earth orbit and promised us a rocket that will be delivered any day now that will be better than the ones we built in the 60's! And people wonder why everyone thinks the government is so inefficient and useless.
True, but phase drift is incredibly uncommon in inverted power systems due to the fact that the inverter can in practice make whatever frequency it wants and it is completely trivial to lock it with the mains, same goes for all other parameters except for power delivered which is entirely dependent on sunlight.
These piratical problems are actually huge problems with syncing power delivered by rotating equipment (gas turbines, diesel generators) etc which take time to react with line changes, and islanding events are quite common if the mains power and the energy load isn't reasonably steady. Not so much of a problem for Solar PV and any consumer off the shelf mains inverter already has the ability to seamlessly take care of this.
That's the only reason why I said elec101, because you can literally buy the components off the shelf and hook them up with no further engineering problems :-)
Most new (commercial) roofs are white, reflective (silvery), or at least light colored, and are often required to be so by energy codes.
I have solar installed and it's already making me a 15% return on investment, so I call BS on your calcs. Go redo them.
My panels are guaranteed to still provide 90% capacity or better in 10 years, and 80% capacity or better in 20 years - so yes, over time, they will be running at a fraction of their new capacity - a very large fraction!
Since installation 6 months ago of 3kw of capacity with a 2.8kw inverter, at a cost of $9600, they have already generated 2409 kWh, with a value of between $505 and $1253. If I used all the power, it would have saved me $505, @$0.21/kWh, but if I don't use the power and it's sold onto the grid I get paid $0.52/kWh. In practice, I have actually used about half the power it produces during the day. with the rest getting sold on to the grid, which more than makes enough to cover my night time cost. I used to have an electricity bill of about $400/quarter. On my last bill, I ended up about $120 in credit.
I get paid a higher rate for power I put onto the grid because it of an incentive scheme put in place by both the govt ($0.40 / kwh) and the power company. ($0.12/kwh). It's helping saving the electricity company from having to build additional power stations to meet daytime peak demand.
Seriously, if you have 10k sitting in a bank or poorly performing investment, you should look into what rebates your local,state or federal government provides, find out what the local buy rate is for solar power and check out if it's worth while for your area. It certainly is in mine.