Datacenters selling power back to the grid? Don’t bet on it, say operators The idea of datacenters feeding power back into the electricity grid during peak demand may sound promising, but operators say it's unlikely to catch on beyond a few trials in Ireland because of the cost and technical complexity involved. So-called "grid-interactive" datacenters use battery energy storage systems (BESS), …
Datacentres, having strained the grid to breaking point, now float the idea of heroically feeding energy back - from emergency batteries meant to keep themselves alive. The scheme? Buy low, sell high - by charging batteries off-peak and dumping the juice back during shortages, like some kind of ESG-approved scalper.
But the whole thing hinges on a tariff system the grid operator can rewrite with a keystroke. One clause change, one regulation tweak, and the grand plan turns into a glorified UPS test. It’s not a solution - it’s virtue signalling so daft, it needs its own backup power.
> from emergency batteries meant to keep themselves alive
There is a certain amount of greenwashing PR bullshit here.
Datacentres would never start discharging their batteries to the grid until they have completed the transfer to Diesel or Gas power. So pretending that it is somehow "green" because it is "from batteries", is bollocks really.
There are already DCs (e.g. Microsoft) which overprovision backup generation by a factor of two or more (Diesel and OCGT, each able to supply the full load of the datacentre, plus batteries for the changeover time) and they intend to run both Diesel and OCGT so that they can export power to the grid when it is profitable to do so.
The bottleneck is the grid connection itself, and they are trying to optimise the use of that by running it both ways (instead of simply disconnecting and falling back to backup generation when the grid pays them to do so, they put their generators into overdrive and their substation into full reverse, because the grid i.e. the tax/bill payer, are paying them top dollar)
The main reasons that this is needed profitable, are 1) because NIMBYs object to pylons, so there are transmission constraints, and the UK is considering having 'locational pricing' instead of improving infrastructure, 2) because the distribution networks are decrepit and there is not enough capacity in dense urban areas like West London, and 3) because datacentre builders refuse to build in places like Scotland where they would reduce grid constraints rather than exacerbate them.
1) because NIMBYs object to pylons, so there are transmission constraints, and the UK is considering having 'locational pricing' instead of improving infrastructure
Ok I might be talking bollocks here, but wasn't there an assessment that underground cabling would be cheaper and less disruptive long term? Seems like the whole idea of pylons makes no sense, unless you are a manufacturer of pylons.
And with conspiracy tinfoil hat on, pylons serve another role - strong local resistance, so they give ministers opportunity to have someone to blame when project is not delivered.
Er, unfortunately no, not really!
Underground cables only make sense for short distances due to the capacitance of the cable (not to mention the expense) see: https://xcancel.com/EngineerLondon/status/1791012963945488815#m
The capacitance (due to the close proximity of the 400kV live conductor to earth) causes excess current in the cable and a phase shift, which also needs expensive equipment (static VAR compensators) to correct.
HVDC is not affected by this (and that is why all long underground cables are HVDC) but HVDC has its own problems - it is not "synchronous" with the AC frequency and makes the grid more vulnerable to "islanding" where one part of the grid ends up out of phase with the other - a headache for transmission operators. It's also very expensive (requiring big, power electronic convertor stations) and tends to be point-to-point only, not in a network.
I'd say that the opposite is true: Pylons are cheap - underground cables only make sense if you are a manufacturer of cables and HV power electronics
But do link to the assessment that you mention
It's not just the capacitance. Under ground (or under water) cables need insulation that overhead cables generally don't, and they also usually need active cooling. Thus the construction of an underground cable is hugely disruptive, even compared with pylons, and they can be more expensive to run and maintain.
There is also a huge problem with what the public thinks of when you say "pylons". For example, someone I know lives near a proposed pylon route taking power from the substation which connects a windfarm to a distant grid connection point. As far as they are concerned, the company involved is proposing 60m tall towers carrying six cable bundles across the hilltops and visible for miles around. The sort of thing you see for 275kV or 400kV lines. In actual fact the company is proposing a 132kV connection, which can be delivered on much shorter pylons, mostly in the valley floor with some (not much) underground, and they've even compromised and downrated from a dual circuit (six cables) to a single circuit (three). Yes, of course it's going to look "different" in the landscape but it isn't as bad as some people make out.
M.
Another problem with underground ACHV transmission cabling is that of expansion/contracting of the cables. Underground HV conduits are oil filled to control this flexing and have there own issues with build/maintenance quality. And when it all goes to shit you don't want to be any where near it
Yeah interesting point. Applies to HVDC too, presumably. How do they cope with expansion lengthways for a direct buried cable? Put loops in along its length?
Also worth noting that for underground transmission, Gas Insulated Lines (GIL) are a thing. This is a wide (~0.5m diameter) stainless steel pipeline, with an aluminium conductor pipe in the centre, surrounded by SF6 or similar insulating gas. They have a fatter dielectric so less capacitance, but are horrendously expensive and require a very wide trench to be dug (10m wide for 6x gas insulated lines, or 15m wide for 18 cables carrying the same current), and takes 3-6 months per km to install (see link below)
@elsergio I wonder if this is the assessment you were thinking of?
It says (p9-10) that overhead line has 70 times less unavailability than cable (because pylons are much easier to repair) and 15 times less unavailability than GIL (but personally I am skeptical of the stated reliability and mean time to repair of GIL, it being a very new technology compared to overhead lines and cables - I can imagine a fault on a direct buried gas insulated line taking a horrendous effort to repair, if the line leaked, arced and became contaminated) also very nasty for the environment if they are filled with SF6
And on p17 it says that cable costs between 12 and 16 times as much per km than overhead line. GIL is claimed to be slightly cheaper than cable for this project (Hinkley-Seabank), but again I am skeptical about that.
The best option by far IMO is Pylons, they are far easier to repair when damaged, and have in my view LESS impact on the landscape, given the motorway-width trenches that have to be dug for both GIL and cable, compared to the small footprints of lattice towers
Er, it isn't...
Batteries can't be used to plug wind lulls (dunkelflaute) because we simply can't make enough of them - we'd need almost 1TWh to power the UK for a day, and the entire global supply of batteries is 3TWh/year, and we can't scale up production by an order of magnitude or three, because the production of batteries is already causing massive (and unacceptable in my view) environmental loss/damage, and they also have a serious waste/recycling problem.
Grid batteries are only used to make money by arbitrage (like with interconnectors) and to keep the grid frequency stable during sudden changes in supply/demand, while we warm up or cool down some gas plants (the combined cycle part of a CCGT is a steam turbine, and it takes a while to ramp up and down) - batteries can plug that half-hour gap, but they can't cover weather-induced gaps without gas.
The cycle round-trip efficiency of battery storage system is 90-92%, so 8-10% is lost.
They work on a small scale (rooftop solar etc) but at grid scale, it's daft IMO. If we're to have any hope of decarbonisation (and have a chance of survival past the last drop of oil), we need to loosen the regulatory stranglehold on Nuclear. (and stop wasting electrons on bullshit machines)
what total and utter tosh
Are you really saying that you think that there is a need to back the whole energy consumption of the UK for 24 hours? why do you think that?
Batteries (and in this case I also include Pump storage Hydro) are great for smoothing out peeks and troughs in demand, and only need last long enough to spin up another generation source...
As for energy efficacy charge vs discharge. take a look at the energy efficiency of a transformer, a linear power supply, any form of heat based power generation system etc.
No, i'm pointing out the enormous gulf between batteries and a feasible "LDES" (long duration energy storage). If we wanted to get rid of gas (and Biomass, which is even worse than gas or even coal IMO), we would need LDES of this order of magnitude. Batteries just aren't suitable for this.
> Batteries (and in this case I also include Pump storage Hydro) are great for smoothing out peeks and troughs in demand, and only need last long enough to spin up another generation source...
I literally said this.
> As for energy efficacy charge vs discharge. take a look at the energy efficiency of a transformer, a linear power supply, any form of heat based power generation system etc.
Of course. But the entire UK transmission network has an overall average loss factor of 1.5%, including all of the transformers, overhead lines, cables etc. between the generator and the grid supply point substation.
Pumped hydro is even less efficient than batteries, at 70-80%. It -is- of course good for LDES, but you can't build them just anywhere, you need an existing mountain
> and only need last long enough to spin up another generation source...
And as pointed out those "other generation sources" will have been decommissioned and hence not available unless we keep them on standby...which will cost a shed load of £££ so kiss those cheaper electricity bills goodbye
The difference is scale. If I have done my internet searches correctly, the UK's largest battery has a capacity of 266MWh and a power output of about 150MW. Compare this with a relatively small pumped-storage station such as Ffestiniog, which can store over 1.1GWh (that's 1,100MWh) and generate 360MW. Once built, the major components last for decades (Ffestiniog was first opened in the 1960s) while batteries, erm, don't.
A mere stone's throw from Ffestiniog, the UK's largest pumped storage station is of course Dinorwig (at least, I always understood it to be the largest). It stores over 9GWh of energy and can produce over 1.7GW at full whack.
The main downside of pumped storage is that while very much faster than thermal or gas plant to get running, they are still several orders of magnitude slower than batteries - tens of seconds rather than tens of milliseconds.
It'd be interesting to see a construction cost comparison per MW and MWh, and a similar lifetime cost to include maintenance. I'm sure the figures are out there somewhere but I haven't seen them.
M.
Yes, as Martin says, the difference is scale.
Batteries are very efficient and very powerful, but they are extremely expensive per unit energy. We need Copper, Graphite, Lithium, Cobalt etc. to make each unit of storage, (which are mined and processed at huge environmental cost) and they have a limited lifespan - the more batteries that we have in service, the more we need to manufacture and dispose of per year just to replace them when they reach end-of-life. It's not sustainable.
Whereas pumped hydro absolutely IS scalable - because gravity is free, Water is abundant and essentially free, and all you fundamentally need to increase the energy storage capacity of a pumped hydro system is more water and/or more elevation. And it never degrades over time.
Wind and pumped hydro COULD work.. If we relaxed planning rules and sold a few mountains out of the Crown estate.. We could make a nice quarry on the top of Mount Snowdon, get some lovely stone for buildings, hollow out the summit into a reservoir, dig some nice tunnels, erect lots of pylons all the way to Scotland, and then we'd have a few more GWh of storage. Of course, a lot of people would object to this destruction of the natural landscape, and they would much prefer to (not) see cobalt mines in Africa, Lithium mines in Brazil, pollution in China, even if they were 1000 times worse for the environment per unit of storage.
Nuclear works though. Yes it's expensive, but only artificially so, due to utterly barmy regulations.. And it is dependable, zero carbon energy! Wind is zero carbon but not dependable, and Gas is dependable but not zero carbon. There's a huge value gap (equal to the astronomical cost of building sufficient and reliable long duration storage) between having one and having both, which many people seem to neglect..
Imagine if all renewables had to implement storage, and all fossil fuel and Biomass burners had to implement Carbon Capture and Storage such that they emitted zero CO2 - that would make nuclear look cheap even with the current level of radio-paranoia.
Interesting couple of points there.
Regarding pumped storage, there are plenty of locations where small- and medium-size stations could probably be sited without too many environmental objections (NIMBYs aside). Don't ask me to name any though! However there is a cost implication so the calculation needs to consider cost and reliability of supply. I'm told that it's near impossible to make a pumped storage station run at a profit when taken stand-alone, so privatisation was bad for this aspect of the grid and whole-grid planning possibly works better.
A bit like bus services.
Back in the 1960s / 1970s there were plans for two stations on the scale of Dinorwig. The second was to be on Exmoor somewhere. Not entirely certain why that was binned, but possibly because gas plants were coming on line, which are much faster-acting than the coal and oil plants which were the bulk of our generating capacity back then so there wasn't so much need. I wonder why this idea hasn't been revived.
Regarding requiring renewables to implement storage, there is some movement on this front. Wind in particular often runs at a surplus and the grid has to issue orders to switch turbines off, or "dump" the power. There is additionally the problem that Scotland produces far more energy than it has capacity to export to the rest of the UK. An HVDC cable is already running down the west coast and there are plans for two (I think) more down the east coast, but still some companies are looking to store surplus generation. Batteries could work for this but a possibly more scalable alternative is to use the surplus generation to electrolyse water and store the Hydrogen. Hydrogen could potentially be pumped into the gas grid but I think the more common suggestion is to store the Hydrogen for later use in fuel cells or in gas turbines to generate power when the wind isn't blowing.
M.
Datacenters as dispatchable load curtailment may make sense. That would work on a facebook or Google scale. The way it could work is if a grid operator sees that an area is overloaded they request the big data center to shed load by transferring workloads to other regions.
Datacenters could be good for the grid, at least locally. I live in a fairly rural area. Electricity should probably cost me at least 20% more per kWh than it does, based on the distribution infrastructure required to feed power to us. Fortunately for me, we have large mines and paper mills around here. The local power utility is built to serve them. Commercial and residential users are small potatoes, so our costs are relatively cheap. I could see datacenters having a similar effect.
Except that telling a datacentre to reduce its consumption actually means telling it to start its Diesel generators - hardly 'green'.. They aren't going to be shutting down or throttling any of their customers when they could just burn some oil instead.
They are NOT good for the grid (if anything, they are quite destabilising) and thanks to soon-to-be-introduced locational pricing, high-demand areas will pay more, so you will be paying that 20% sooner or later.
(didn't downvote you though)
Datacenters could be good for the grid, at least locally. I live in a fairly rural area. Electricity should probably cost me at least 20% more per kWh than it does, based on the distribution infrastructure required to feed power to us. Fortunately for me, we have large mines and paper mills around here. The local power utility is built to serve them. Commercial and residential users are small potatoes, so our costs are relatively cheap. I could see datacenters having a similar effect.
This touches on another element from the article-
"The problem is that they [nuclear plants] are not located where we need them. In France, we need them in Paris and Marseille, and instead they are in the countryside, not exactly where we need them. So either you do that, or you try to manage to produce your own electricity, which would mean that we would become an energy company."
I rather doubt a paper mill or mine needs a terabit fibre connection. A few bit-barns might. I also very much doubt a lot of bit-barns actually need to be in Paris or Marseille, especially not most AI datacentres that generally aren't latency sensitive. So you could get anywhere in France in <10ms, and might even have latency benefits by shifting workloads closer to France's borders with the rest of the EU. Which is also why Dublin never really made much sense (other than tax-wise) given it's one of the farthest flung corners of the EU and a lot of traffic from the US ends up taking a loop around London, Paris, Amsterdam etc before hitting Dublin.
Which is also one of those joined-up-govermnent challenges. Bunch of European nations all wanting to be the AI centre of Europe. Bunch of Eurpean nations struggling with energy demands (self-inflicted). No national policy (yet) to say "Build it here!" where there's a power surplus(ish), far cheaper land than Paris or Marseille. And then as part of this cunning plan, run 288f cables accessable on a FRAND or cost-plus (or even minus) to existing hubs like Paris or Marseilles. Or the landing stations. Or even French borders and create... err.. EU Net and call it good. Design-wise, it's made simpler because OSPF just means following the power cables. Things like wayleaves might become a bit of a PITA (as always), but being a government, one can generally make those problems go away.
Then DC shops like DR can shunt latency insensitive workloads out of Paris or Marseille, providing latency stays in contract with their customers SLAs. And no need to try faffing around with firebricks.. I mean batteries.
Dublin built a huge fiber ring around Dublin when building the M50 orbital ring (semi circle?) so there was tons of dark fiber waiting to be used in nice big conduits with lots of spare capacity. So when MS, Facebook et al were looking for sites you could build a DC anywhere near the M50 and have (at the time) fantastic connectivity without a lot of extra work. It was also linked into the IFSC in Dublin which had a lot financial related companies (Charles River, FactSet, JP Morgan etc) and this was kicking off in the late eighties for the IFSC and the late Nineties for the T50. Around the same time fiber was laid from the west of Ireland and Cork to connect to the T50 and elsewhere so ingress in the west and south from undersea (at that time) was easily available. Later in 2015/16 a 52Tbps fiber connection from Co Mayo in the west of Ireland direct to Long Island was completed but had been on the books since 2005 but 2007/8 crash halted any real progress on anybody even considering moving ahead.
Add in English speaking educated workforce, access to the EU, US friendly government, favorable taxes temperate climate for reasonably fixed cooling costs and a national grid that was if not excellent at least stable and it was a slam dunk. No one of those factors would have made the deal but in combination it was very attractive to the big data center players. And still is
Dublin built a huge fiber ring around Dublin when building the M50 orbital ring (semi circle?) so there was tons of dark fiber waiting to be used in nice big conduits with lots of spare capacity. So when MS, Facebook et al were looking for sites you could build a DC anywhere near the M50 and have (at the time) fantastic connectivity without a lot of extra work.
Yup, but practically that work wasn't easy. So Eircom took a beating from previous owners, and very tightly regulated by ComReg. Which made exploiting any of that fibre a tad tricky, ie the usual regulatory and incumbent shenanigans getting access to duct sharing, DF or even wavelengths. Plus some.. interesting decisions from DCs. So MS partnered with Verizon for 'foreign' network connections, making it a tad diifficult for non-Verizon customers to build high capacity connections into those sites. MS seemed to expect business users to be happy with public IPVPNs, or if they wanted private, Verizon's private IPVPN.
If you wanted a 10Gbps Ethernet, tough. Which can be a common problem with some datacentres, ie tied connectivity that doesn't always suit large customers. But then those datacentres often don't want to try and manage lots of 3rd party connections, especially if they're service oriented like MS, Oracle etc.
Later in 2015/16 a 52Tbps fiber connection from Co Mayo in the west of Ireland direct to Long Island was completed but had been on the books since 2005 but 2007/8 crash halted any real progress on anybody even considering moving ahead.
Yep, good'ol Hibernia Atlantic.. Who did an.. interesting move to offer 'premium' low-latency connections to traders and artificially increased latency to the non-premium users by stacking drums of fibre. But then got caught up in GTT's spending spree, which prompted GTT to go titsup.com and their infrastructure flogged off to this interesting outfit-
https://en.wikipedia.org/wiki/I_Squared_Capital
In March 2016, I Squared bought the Irish power company Viridian for about one billion euros. Viridian provides about 20% of the power to Ireland through gas-fired power stations and wind farms.
Which meant there is/was a company that could provide both power and bandwidth in Ireland (and some other countries).. Although I Squared then span out it's network assets into a subsidiary, EXA Infrastructure. All part of the joy of telecomms, but EXA's a good company to work with, if you need bandwidth.
But there's still the issue of geography, and Dublin's distance from staff or customers if the majority of your business is within Europe.
"It requires different equipment and different investments to be done up front,"
A bit like the London Sewage system then.
From Wikipedia, "During the early 19th century the River Thames was an open sewer, with disastrous consequences for public health in London, including cholera epidemics. These were caused by enterotoxin-producing strains of the bacterium Vibrio cholerae. Although the contamination of the water supply was correctly diagnosed by Dr John Snow in 1849 as the method of communication, up to the outbreak of 1866 it was believed that miasma, or bad air, was responsible.
Proposals to modernise the sewerage system had been made in the early 1700s but the costs of such a project deterred progress.
Further proposals followed in 1856, but were again neglected due to the costs.
However, after the Great Stink of 1858, Parliament realised the urgency of the problem and resolved to create a modern sewerage system."
Everything costs more initially than throwing your shit out of the window or burning the planet, doesn't mean it shouldn't be done.
> Parliament realised the urgency of the problem and resolved to create a modern sewerage system."
Sadly they did not do it right as both the household waste water and rain/storm water still goes down the same pipe which overflows during heavy rain, hence they bodged it with the London Tideway Tunnel
Ideally the two sewer networks need separating (I think new housing developments as built this way but still connect to a single downstream sever) but the cost would be staggering and we are already in the UK being fleeced for more £bn to provide better bonus's and share dividends prevent raw sewage spills which the companies should have already done.
Bazalgette did his best. Given the space constraints (two sets of sewers would have been vastly more expensive and disruptive) he oversized the sewer by something ridiculous like three times to take account of population growth, further urbanisation and thus more rain run-off. It's lasted remarkably well. Rainwater in sewers is not necessarily a bad thing as it helps transport blackwater and dilutes it which can make it easier to process at the end of the sewer.
In terms of current practice in the UK, surface water from large housing developments is generally allowed to run into the normal sewers but nearly always requires a Sustainable Urban Drainage Scheme. As a minimum this consists of an agreed maximum release rate (i.e. into the sewer system) and sufficient storage to deal with rainfall. Many new developments are incorporating surge ponds or other such schemes which might also allow rainwater to percolate into the ground. There are new rules about paving too, which mean that many driveways are now built with permeable surfaces which again reduces run-off. In general the rule for individual new builds seems to be "no", so other solutions have to be found.
As a private homeowner in the UK, if you can prove that none of your surface water enters the sewer (for example you have a soakaway or can run-off into a stream or other watercourse) you get a discount on the sewerage part of your bill. Note that if you have a driveway sloping down to a public road this can disqualify you unless it is made with permeable paving or drains into your soakaway.
We have a rainwater recovery system where water from the roof (not from the ground) feeds into a tank and is then pumped into the house to fill the toilets (and potentially the washing machines). Excess overflows into a stream at the rear of the property. Although a lot of rainwater does actually end up in the sewer, this system still qualifies for the discount. Maybe they also take account of the fact we use less mains water.
M.
I think the key point was the pricing structure from eirGrid
Looking at this from a bean counter perspective, if it's going to cost I dunno 1 millliiiiiion euro to install battery based UPS backed by diesel and NG generators on top of that. Now if I over provision and install 10 milliiiiooon euro of batteries but can arbitrage the grid pricing to charge the batteries at low cost and export at a high price my amortization of option 1 is say 200K a year but option 2 is 2mn a year (tax benefits there) and I either break even on costs or make a profit.
Then I can see that being attractive to a datacenter owner, not for any social responsible reason, just bottom line drivers. And if your DC is in an area with a less favorable time of use/export pricing then the numbers don't stand up and you stick with option 1.
On demand surge pricing battery supply to the grid seemed to generate seriously good profits in Australia, but I have no idea how their pricing structure relates to Ireland or anywhere else.
The few datacenters that my company has stuff at - they don't use batteries to bridge the gap between utility power loss and the gensets taking over, they use flywheels instead.
They literally only store a couple of minutes of power, tops.
If my limited dataset is indicative of other datacenter designs, might be why this really hasn't gotten off the ground.
But who knows.
The idea fails a basic look at the premise.
When is the battery storage at a data centre going to be used? When it can't get electricity from the mains.
When would it make sense for a battery storage facility to feed power back to the grid? In exactly the same scenario.
So rather like with solar power, they feed energy back when it's not needed. When it's needed they can't feed it to the grid because they need it for themselves, which is the reason they built it in the first place.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
