Your data centre UPS could feed power to the smart grid, suggests research Data centre operators could deploy uninterruptible power supply (UPS) systems which link up with the electricity grid so as to increase its reliability by smoothing out the unpredictability of renewable energy resources, according to research firm Omdia. The research, which is bound to raise a few eyebrows in the risk-averse …
Just like the articles about how car batteries that are idle at home can be used to power shortfalls in grid supply I have one pressing question.
What happens if you need all that power that should have been in the battery but is now reduced because it is been syphoned back into the grid to make up for a shortfall?
Nobody appears to be able to address this. A UPS is generally sized for a specific load and runtime. That needs to be available at all times to cover an event where the power is lost. What is the use of a UPS that has the runtime reduced because it has been exporting power to boil kettles in the advert break or because the wind is not blowing enough?
You can see the conversation now
CEO: We specced and paid for UPS capacity to cover a 2 hour outage, why did it fail after 80 minutes?
Eng: Because Clive signed up to exporting our "excess" to the grid - it's only supposed to use about 10%, but turns out the cells have cycled a lot more than we expected.
"What happens if you need all that power that should have been in the battery but is now reduced because it is been syphoned back into the grid to make up for a shortfall?"
That's a great question. If you don't have an intelligent system managing things, the answer is probably "you're screwed".
If you do have an intelligent system managing things, the answer is "you wont".
Let's use a (admittedly very simplified and idealistic) vehicle to grid example: suppose you run a fleet of battery-powered beer trucks, and you have them hooked up to a V2G system. It's late afternoon on a hot day, and your fleet is at the chargers, ready to charge overnight. Meanwhile, everyone's AC units are kicking in across the city and the local utility is having problems coping with the load.
There are two big possibilities:
1) the grid operator's system asks your fleet (and others) to provide some VARs to help frequency stabilization for about 30 minutes while additional capacity is brought on line and some loads are shed. Your fleet resumes charging later in the evening.
Or: 2) no V2G is available, and you have a major blackout. Your vehicles don't get to charge overnight, but that's ok, because none of your customers are open the next day, so there are no deliveries to be made, and your plant is dark so no more beer is being bottled.
I'm skeptical of using UPSes for the reason you mentioned, but I suppose the same concept could apply.
The situation with a UPS is you have to assume you're going to need all of it starting a second from now.
As to your fleet example, a major blackout might strike about 3 in the morning and only affect the area round the depot. That doesn't mean your customers will also be affected. You're still going to want the fleet fully charged. Your sales and customer service departments aren't going to be happy if the bean-counters did a deal without telling them.
There's also the efficiency of replacement to consider. At a dedicated large scale grid battery substation they will be swapping out the same number of old batteries for new batteries ever week, and the layout will will be designed around doing that in a cost efficient manner - most likely with a forklift truck. That won't be the case the UPS in the data center. They'll be carrying by hand and trolley.
And lets not talk about how much it would cost that Telsa owner to replace the battery.
I could see where the backup gensets could be used for helping during a brown out in the grid but I can't see how this would work for UPS as they only have short run times to give the genset time to start. This also has the issue that the UPS and gensets are sized for the load and if a brown out happens or worse a total power loss then the data center will want all the available power.
This does have the feel of a report done by people who don't live in the real world
What this shows is desperate people clutching at increasingly feeble straws.
Gridwatch suggests that electricity demand is at ~42GW load at peak.
In terms of generating capacity we have:-
30GW of gas turbines
7GW of nuclear (rarely all online at the same time due to maintenance and refuelling)
4GW of (mostly mothballed) coal
3GW of coal plants converted to burn trees
= 44GW.
So bluntly, excluding imports and wind we have just enough power generation to survive one nuclear plant doing maintenance or refuelling and then we have a blackout.
We do of course have 25GW worth of wind turbines, which should be supplying over half of our energy needs. Unfortunately, the graph showing the feed in for the last year on gridwatch shows those actually generated about 5GW worth of power for something like half to two thirds of the year, and 10GW in the remainder.
Which means that the prevailing energy policy has completely and utterly failed at generating electricity (although it has generated good money for jobs for the boys the green industry by increasing the price of the power that is actually generated) and we are now left staring at the probability of blackouts as old capacity with actual power outputs is decommissioned and replaced with replacement wind turbines with "up to" power outputs that everybody but the terminally delusional knows will never actually be reached.
In terms of lost capacity, Dungerness B (1.3GW) closed last year, Hunterston B (1.2GW) has just started decommissioning , Heysham 1 (1.3GW) goes in 2024, Hinkley point B (1.3GW) goes in 2024 and we are committed to decommissioning coal by 2025 (3GW) which is 8.1GW of generating capacity gone in the next 3 years with nowt but wind turbines and prayer (to mother nature that the wind will blow continually) as replacements.
Hinkley point C might come online in 2026 if it's on schedule with 3.2GW worth of juice, meaning we'll be down 4.9GW worth of capacity.
And of course, people are being asked to decommission gas boilers for electric heat pumps and go with electric cars, so demand is going to increase over this time period.
How do you deal with that? Well, this suggests that asking everybody in the country with backup gensets UPS's to feed in supply when the grid is desperate is the next plan.
@Peter2: Thanks for the very well structured observations above. For a long while now, the UK has faced increasing demand, with decreasing indigenous supply capabilities. IIRC, the UK has been heading towards 'negative equity' in it's electricity generation capabilities for a while now, and we're usually not a long way away from running short - as Peter2 points out, we only have to suffer an un-planned outage at a relatively small capacity generation source before we get to the the decidedly nasty 'low frequency' situation.
Yes, there are functioning 'interconnector' cables between us and Europe to facilitate the import or export of surplus electricity, and I believe more are planned to increase capacity. The cynic in me suggests these only exist as it's possible to make money in the very dynamic and complex wholesale electricity market that is evolving (Has already evolved?), rather than being there to benefit the consumers directly with a better security of supply.
Either way, and even with expansion of the bulk interconnector network, it's hard to see these making up for the significant 8.1GW (Ref:Peter2 above) upcoming losses, not to mention the continuing loss of generation capacity that remains ongoing.
The really interesting thing for me is the transmission of, rather than the generation of, the increased electricity usage that is upon us now - just last week, the SMB I work at installed a dual-outlet EV charge point, which (If the right combination of C-Suite vehicles are connected) draws an additional 64A*, or about 14 to 15 KVA of additional load. As more and more people connect EVs, the low-voltage distribution infrastructure is going to face significant challenges carrying that power increase. I really don't want to see an increase in outages and even cable fires as the aged infrastructure, most of it underground, is asked to deliver way more juice than it was ever designed to handle.
The sooner that the boffins (A term of endearment!) get commercially viable Fusion power going, at scale, the better!
*That's 32% of the capacity of the fuse in that phase's supply cable in one go! At least I think it's 32%....I hope the unlabelled fuse carriers do hold 200A fuses and not the more domestic 100A ones! Also, please don't get me started on the phase imbalance that this has created internally and/or externally with the other 2 phases we have!
We are on a medium sized ind est in the South West and are looking to add plant so contacted Western power about upping the fuses to our unit. The answer came back no available spare capacity in the area to be had, not even how much do you want to spend.
It's a bit worrying as there are at least 3 big housing builds going on around us, where will they get supply from?
The housing development does not need electricity, in the same way it does not produce sewage or water run-off.
Then the people moving in have no need for schools of doctor's surgeries. However the builders will have offset some of the environmental damage by planting a hedge or a few trees that invariably die within a year or two. This is what is needed to meet the planning requirements, nothing about the actual infrastructure that has to support it.
This is repeated up and down the country.
Sorry, feeling very cynical about this after going to a Q&A session for a local housing proposal on a "brown field site" that has been growing crops for years. The only time it is brown is when it has been ploughed. Apparently there is (was) some old building on one corner of the land that means the entire site of 3 fields is not "green field".
Fusion hasn't finished being developed. It's likely to be the long term solution to humanities power generation requirements, but it's not an immediate prospect.
We need to be building things right now (as in today) to keep the lights on in the next 5 years which means things that have been developed and actually work here and now. Drax is apparently going to build a new gas plant to replace the coal capacity lost, however this is subject to legal and physical attacks from increasingly delusional green types who despite the obvious issues pointed out above say that it's not required, and we should spend money on batteries and wind turbines instead.
As above, wind turbines aren't working. You can see the math on batteries on another post below, but 1GW of battery storage for 24 hours would cost approximately £96 billion. It'd also last ~5 years until the batteries would need replacing and 1GW storage capacity for 24 hours is in any case far short of any useful requirement.
96 billion quid would for general context be enough to build 5 nuclear sites of the Hinkley point design outputting 3.2GW each so would generate 16GW total. A glance at the graph on Gridwatch for the last year will immediately show what impact this would have on us burning fossil fuels; simply put our gas generation for electricity would fall to occasional spikes of 5GW and we wouldn't be burning *anything* for at least 7 months of the year.
Personal follow-up, months later!
"hold 200A fuses and not the more domestic 100A ones"
[1] New EV for one of the guvnors...non-hybrid/full EV vehicle with increased charging power requirements.
[2] Sudden un-solicted loss of one phase, which takes out a good portion of the building
[3] Traced back by yours truly to a blown fuse on the supply side of the meter
[4] Power company turn up to replace the fuse...I inquire about phase balance.
[5] Reply:L1 drawing ~ 5 to 8A. L2 drawing about 7A. L3 (repaired) drawing 70A nominal,>82A peak.
OK, 82A...unbalanced but not disastrous...right up to the point where we learn that the fuses were only 60A.
UK Power Networks couldn't have been more helpful. Supply upgrade requires new fuse carriers in tn the DP/Meter area, and external underground works it seems.
The typical small scale UPS kicks out a Kilowatt. A Megawatt (the next unit up) is a thousand kilowatts. A Gigawatt (national grid scale) is a thousand megawatts. Therefore, to back up one gigawatt for one hour, you'd need 1*1000*1000= one million killowatt scale UPS's.
At ~£400 each this would cost about 4 billion quid, exclusive of a building to put them in and the power supply stuff. So a gigawatt worth of storage for 24 hours would cost £96 billion quid, and would need replacement in 5 years. (which is the battery lifetime)
In comparison, building a carbon copy of Hinkley C would cost ~£20 billion and would generate 3.2GW for 60 years, with possible lifetime extensions if it's in good condition in 2080.
Grid operators don't "invest" in battery storage because everybody can see that it's a stupid idea.
Provided it means there will always be enough power for the UPS to do its main job.
Octopus Energy in the UK have already teamed up with Tesla to provide a VPP (Tesla Energy Plan), with consumers getting charged a much lower rate for electricity for joining the program, so it looks as if everything is in place to make this happen.
If I have PV and a battery bank at my house then I would expect to see a much larger saving than that page is suggesting.
With the right setup you shouldn't need the grid at all, trouble is this would not work every type of property.
In most of the world, if you have PV and batteries there's no reason to be on the grid, full stop. It's more of a liability than an asset, especially with the utilities lobbying for "grid participation fees" in so many places, which of course they would "invest" in higher executive bonuses rather than grid resiliency. The times when you'd benefit from it are the times of peak demand, summer heat waves and winter storms. Those are exactly the times when the grid fails anyway. It was a good idea when most generation was fossil or hydro and most demand was baseline; those mechanisms benefit tremendously from economies of scale. Fission is another matter; it has huge diseconomies of scale, a story for another day. But PV and batteries scale pretty linearly and prices have dropped -- assuming you don't do something silly like buy lithium batteries -- to where you're coming out pretty much even with utility pricing. If you're getting tax incentives or want to factor in future inflation, decarbonising costs, and politically-driven utility rate increases, you're probably going to be ahead. So if these technologies make sense at your location, you may as well go all the way. The grid's relevance will eventually be limited to high latitudes and a subject of intense "what went wrong?" debate in business, economics, and public policy.
As above not only does investing in a UPS mean you expect to have a runtime normally specified based on cost i.e. 20 minutes costs X and 40 minutes X*3, big bosses aren't going to be keen on the extra risk.
But if you are charging/discharging then you are using the limited number of battery cycles and X/2 is probably the cost of replacing the batteries.
Both costs will need to be factored into any payment.
a) What's in it for the DC operators ? I certainly wouldn't want to turn around to all the customers who are paying for UPS backed feeds and saying "Oops we gave all the power back to the grid"
b) I was always under the impression that there was less than an hour (if not less than 15 mins) of power in the batteries. They are there to smooth the load before the generators kick in to provide the power after all.
I know a number DC operators have heaters (typically water radiators) on their generators so they can provide load sooner
a) What's in it for the DC operators?
Money, obviously. How exactly that would work would depend on what the utility offers I guess. Maybe the DC operator's purchase/installation of the battery system is subsidized, maybe they get access to a lower rate tariff, maybe they get monthly payments, maybe they are paid as the excess capacity is utilized by the utility. Probably some combination of the above.
That would allow them to get a large enough UPS that would maintain the amount of battery backup they require (which only needs to last long enough to get the generator running, they are not designed for long runtime) with the excess capacity used for the grid's peak shaving / excess renewable energy storage use while still lowering TCO below what it would have otherwise been.
Money, obviously
I strongly doubt that there will be enough money in it. The floor space is there for the servers.
* Batteries are not cheap otherwise they there would be ALOT more companies already competing in this space.
* The floor space is better used for other things, like the routers, switches, servers and disk arrays the DC operators are already selling to.
* Data centres don't have loads of extra UPS capacity. If my servers go offline due to a power outage I'll be demanding compensation and there's a strong possibility that I'll be terminating the contract for a different hosting provider. I think most customers of hosting providers would have the same view,
As I said earlier most DC UPS's are designed to fill the gap between loss of the mains and the generators getting up to speed to take the load. While the grid will have spikes where extra power is needed most of the time there isn't spare capacity for hours after. That's why pumped hydro and the like are a thing.
Pumped hydro (and similar grid level storage) recharge over night to supply power the following afternoon / evening. That's also where the power is the cheapest so they can make money. They don't supply energy at 5-5:15pm then recharge at 5:30pm. Go take a look at Gridwatch.co.uk
The reason why datacenters are ideal for this is because their large power draw means they have high capacity power delivery, and only such locations are suitable for grid connected batteries.
If the utility is faced with a choice of funding 100% of the cost of a battery farm, or could instead fund <100% of the cost to essentially co-locate at a datacenter, they will save money.
If you can get 50x the size of battery you would otherwise install or more for the same price that you'd pay for the "proper" size along with a commitment that the utility will maintain it and provide a guarantee sufficient capacity will always be available for your needs (and often more) it seems a no brainer to me especially if there are additional incentives like a better electric tariff since during peak periods your datacenter can draw from the battery instead of the grid.
> If you can get 50x the size of battery you would otherwise install or more for the same price
But as DevOpsTimothyC has pointed out, your bigger batteries are now taking more space in the datacentre.
You can argue some of the space was "dead" anyway, but not all of it will have been, and space in a room full of server racks is most valuable when used for.... renting out racks.
With your suggested approach
- you still have the capital cost of the UPS (because the utility is only really covering the extra)
- you'll have higher install/compliance costs because of the massively increased capacity of the system (but, the utility may subsidise that)
- Ongoing maintenance will likely be more expensive, but you've said the utility is covering that
- You've got more floor space taken up by the UPS (you don't get 50x capacity without an increase in size), so may have had to remove a couple of racks
Essentially, you've saved on some UPS maintenance costs (which aren't that high for the routine stuff) whilst reducing your source of income. Unless the utility has also significantly reduced the per-unit cost of your electricity, you've made a loss. You certainly aren't going to make that back from payments for what you feed back in.
And that's *before* you get onto the issues with trusting the utility's commitment that they'll maintain it (and ensure you have enough power). If it turns out they cut corners, then at best you get an outage (losing you customers) and at worst you follow the path some of OVH's datacentres took last year.
For the DC provider, there really isn't much of an upside.
"Instead, about 80 per cent of respondents estimated that between 10 and 50 per cent of the capacity of the battery systems deployed in their data centre today is excess..."
First of all, 10% is not a lot of margin. Battery capacity varies with temperature and age and most battery banks at DCs are designed to last only long enough to start generators; 10% of that time is not much. 50%, on the other hand, is a lot of excess margin, but either it was specified that way for good reason (to survive a 99th percentile generator startup time instead of 95th, say) or it should never have been designed that way in the first place and is a waste of money. In short, there should not be "excess capacity" in DC UPS battery banks.
"According to Schneider Electric, these changes have been made possible by newer UPS systems using lithium-ion battery technology rather than the valve regulated lead-acid (VRLA) batteries traditionally used."
This is both true and a lie. Flooded lead-acid (same chemistry, different maintenance design) is no more expensive than VRLA but with proper maintenance -- which can be partially automated -- have design lifetimes up to 5000 charging cycles. There are two basic parameters that affect lead-acid battery lifetime at a given depth of discharge: plate thickness and electrolyte maintenance. VRLA batteries "wear out" because they lose electrolyte during charging and it's not replaced. The valves regulate this loss but do not eliminate it. Because of this limitation, they are also designed with thin plates because there is no point in having plates that will outlast the electrolyte. Thinner plates also allow greater current to be supplied by the cell, which is important for applications like starting an engine, but for UPS applications even very thick plates can supply more than adequate current. Flooded batteries have no valves but are designed to have water added periodically to maintain proper electrolyte levels. They can be, and are, designed for a wide range of points on the curve of cost and max current vs. lifetime and depth of discharge. This is the main distinction in whether the batteries you see at the auto shop are being sold as "starter", "deep-cycle marine", "off-grid power", etc. Installing thin-plate VRLA batteries in a stationary application like a DC where weight poses no problem and a staff is always on site to perform periodic maintenance is insane. A DC UPS battery bank will have a very small number of very deep discharge cycles and will otherwise be maintained at the float voltage, which means electrolyte loss will be minimal. Whether FLA or VRLA, such a bank should have a lifetime of 7-10 years minimum and possibly longer. There are many batteries on the market today with these attributes and warranties of 7 years or more. Lithium-ion is much lighter and more compact -- important for mobile applications but unimportant for stationary use -- and is maintenance-free, but it is also 3-10x as expensive at a given capacity point and comes with a small risk of causing a fire that is all but impossible to extinguish. The only reason to push for lithium chemistries in an application like a DC is because you're selling ignorance. If your goal is to market your DC to people who think lead-acid is "obsolete" and lithium the new hotness, by all means. Or if you're a manufacturer of lithium-based batteries and can convince a DC manager of that on the golf course, again bully for you. Otherwise it's a joke.
'"In addition to voltage and frequency regulation plus power back-up, a data centre UPS can be used for peak shaving, demand response, and even to send excess power into the utility grid for profit," he said.'
Yes, and the equipment suppliers listed have been selling electrical equipment for many years that offers all of those features and support a wide range of battery chemistries, voltages, and capacities. All of this has been on the market for over 20 years and has been widely-deployed with mostly FLA or VRLA battery banks. This feels like nothing but a marketing push: talk up "renewable energy grid integration" and "new frontiers opened by lithium-ion batteries" because those things sound sexy and people read about them in newspapers; maybe it'll sell more electrical gear. In fact there's nothing new about any of this; you could (and many have) built systems that do all of this in a dizzying array of scales, applications, prime power sources, and grid integration or lack thereof. There's no good reason to add these features to a DC UPS unless the DC is powered mainly by local generating sources rather than the grid, in which case the UPS is not really a UPS any longer but a core part of the electrical supply system with very different design goals.
https://fgw.ddnss.de/StromSpeicher.html
An online translator can give you an English version.
Chemical Batteries are in the order of 100x to 1000x too expensive to store a nation's electricity for a month. Also, they wear down in a few years, unlike a concrete construction, which can last 100 years, if properly done.
Batteries have a limited number of charge/discharge cycles. There will need to be massive incentives for it to be worth shortening the lifttime of your battery packs, and those incentives will mean very expensive electricity... More expensive than starting-up a peaking plant.
This is a bit like Uber and all the other contract delivery driver scams. You already have a car, right, so driving someone/something around all day is just like free money for you, right? But there aren't that many big data center operators with so little intelligence, so it doesn't seem like there will be any takers.
But if I (my DC anyway) pays for and maintains a UPS system, I want to have it as available as possible for when I need the "U" part.
If the lights go out and the fans stop spinning then my downtime per minute is going to cost WAY more than what I could potentially gain from selling some kWh to the grid, especially if I cant use it.
There is also no mention of the difference in cost between usage and generation of electricity. Current (capped) consumer tarrif is around 21p/kwh (business tarrifs are much higher) and the best export tarrif I can find is 7.5p/kwh. There are probably better rates on negotiation but I don't see a world where this is profitable.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
