Britain’s Blackout

Three sources of power generation failed within half a second of one another.  6% of generation, or 1.9 GW, was lost.  The grid is only designed to cope with the sudden loss of the largest single component in the power system (normally Sizewell B at 1.2 GW). That could have covered two of these failures, but not all three simultaneously.  Not enough power could be supplied to meet demand, and so 5% of customers were automatically disconnected to prevent the grid from complete collapse.  While this was painful for those involved, it saved the whole country from losing power, which could have taken days to recover from.

The problems began when lightning struck a power line just before the Friday rush-hour.  This sounds dramatic, but it happens thousands of times a year.  Circuit breakers automatically disconnected the line in less than a tenth of a second to isolate the current surge from the lightning bolt.  About 20 seconds later they reconnected the line, and normally this would have been the end of the story.

The large wind farm at Hornsea (100 km out in the North Sea) disconnected itself from the grid 0.3 seconds after the lightning strike.  It detected a voltage disturbance when the power line was disconnected, and attempted to help correct it by injecting ‘reactive power’.  Something went wrong and two banks of wind turbines disconnected to protect themselves.¹ 737 MW of generation was lost.

Within half a second of the lightning strike, a steam-turbine shut down at Little Barford gas power station, which is connected to the affected power line.  The turbine had reacted to abnormal speed readings, which were probably not true because the control systems were disrupted when its power supply switched over to battery back-up.¹ This is unusual, and looks like an equipment failure of some kind.  A further 244 MW of generation was lost.

The voltage disturbance caused by the power line and generators disconnecting then rippled out through the grid.  Lots of small distributed generators² saw this and the resulting drop in frequency as a loss of mains power, and so disconnected for safety reasons.³  These systems were still using outdated forms of ‘Loss of Mains’ protection which are banned in new installations, as it is known they can be outwitted by grid disturbances.  Another 500 MW was lost.

With over 1,480 MW of generation suddenly disconnected, more power was being taken out of the grid by consumers than was being put in by generators.  This caused all generators across the country to start slowing down, and their speed – the grid’s frequency – fell from 50 Hz down to 49.1 Hz over the next 20 seconds.  While this fall of 1.5% might not sound like much, National Grid has an obligation to keep frequency within 1% of its normal value, its “statutory limit” of between 49.5 and 50.5 Hz.

While the frequency was dropping, the first line of defence started coming to the rescue.  National Grid keeps “response and reserve” to cover certain types of failure, and was holding 1,338 MW of frequency response made up of batteries, automatic demand response and part-loaded generators.  Over the coming seconds 1,080 MW was delivered, which arrested the fall in frequency.  A second gas turbine then failed at Little Barford and a further 210 MW was lost.

This sent the system frequency below the critical threshold of 48.8 Hz (2.4% below normal).  At this point, National Grid ordered the second line of defence, “Low-Frequency Demand Disconnect” (LFDD).  Like a surgeon deciding to amputate a limb to save the patient, around 5% of demand was disconnected from the grid.  This helped to bring demand back in line with supply and allow frequency to be restored, but it meant cutting power to over a million customers.

THE MINIMUM SYSTEM FREQUENCY EXPERIENCED EACH DAY SINCE 2009

Nothing in the August power cut was a result of wind being intermittent or variable.⁴ While wind output was high that day, supplying nearly a third of demand, there was no sudden reduction of wind speed leading to loss of wind power.

There was a sudden loss of one wind farm but others in the region carried on operating as normal.  Hornsea was only partially built, and not all wind turbines had been commissioned.  Two 400 MW modules were running under some temporary arrangements with the grid, and not in their final configuration.  This is a pragmatic arrangement as it allows partly-built wind farms to generate clean electricity rather than keeping their turbines idle, but this may need reviewing in future.

There is a broader question around the move to higher shares of renewable generation. Wind and solar farms do not have heavy spinning metal turbines which provide ‘inertia’, the shock-absorbers of the power system. As we move to less conventional generation, the power system is seeing higher rates of change of frequency. This is giving National Grid less time to react to problems on the system.

How much time we have to correct a problem before system frequency has fallen by 1%

This ‘speeding up’ is a direct result of more wind, solar and sub-sea interconnectors.  However, it is being counterbalanced by new technologies that provide response and reserve services.  Nearly half of National Grid’s response came from batteries, which were exceptionally fast to reach full power.  The traditional spinning generators, which would have been the only recourse a decade ago, took much longer to respond.

If we want to stop power cuts happening we have to think about all the causes. Focusing on sudden outages of generators, we would need to look at the regulation that defines what National Grid has to guard against: the Supply Quality and Security Standard (SQSS). This requires the grid to remain operational after the loss of a single ‘generating unit’, ‘power park’ or ‘converter station’.  National Grid holds enough response and reserve to cover for losing the biggest unit in operation that day.⁶ 

The obvious question is do we want to cover for two large units suffering problems, rather than just one?  Doing so would have meant that the 2008 and 2019 power cuts would have been avoided, but not the 2003 or 2013 ones.  This wouldn’t cover three large units failing in quick succession.  It would mean an increase in costs of procuring those response and reserve services.

There are many issues to consider and lessons to learn.  First would be to uncover the detailed technical reasons for the outages at Little Barford and Hornsea and try to ensure that both of those problems are made less likely.  Hornsea was only partly complete and running under temporary grid-code arrangements. Perhaps new farms should have grid code compliance assured one module at a time, and operators should be more careful over the configurations that may be used for very large wind farms.

The second issue is that of common-cause events.  The 2008 power cut was two power stations failing in unrelated events – just bad luck.  Should we see this 2019 outage of two power stations and a share of distributed generators as having a common-cause?  They were all unfortunate reactions to the lightning strike, but perhaps this was also in ‘bad luck’ territory It was already known that older distributed generators were prone to ‘sympathetic tripping’, disconnecting when mains electricity was not actually lost.  The protection systems of older and vulnerable distributed generators are being replaced in an ‘accelerated’ industry campaign that completes in 2022.  Perhaps this could be accelerated further?  Perhaps National Grid should factor these devices into their largest loss of load calculation until they have been upgraded?

Finally, the ensuing power cut caused more disruption than should be expected. Newcastle Airport did not class itself as critical infrastructure.  Ipswich Hospital did not cope with the switch to backup power.  Some trains were set to turn off, and many could not be restarted. Organisations across the country would be well-advised to check if they can and should be Protected Sites and whether their own protection systems are appropriately configured.

Then looking beyond technical elements: perhaps there needs to be a public discussion about the appetite for paying to reduce the risk of demand disconnections (LFDD) by requiring National Grid to purchase larger volumes of reserve and response.  Similarly, perhaps this event will raise awareness for critical infrastructure provides about what sort of grid conditions should be planned for and immunised against (e.g. adjusting over-sensitive internal protection, providing UPS etc)

There is nothing to suggest these events are more likely today than they were 10 years ago, so it would be difficult to justify spending more to reduce the risks in future.  That said, Britain’s power system is changing rapidly and so we need to be ready for the different kinds of problems that will be faced in the future.

Both the supply and demand for electricity are going through unprecedented changes as new technologies take hold.  Britain already has more renewables than fossil capacity, and the balance is set to shift further as all coal stations retire by 2025. More wind and solar will come online, backed up by batteries, more interconnectors to the continent and bigger nuclear reactors.  National Grid is preparing itself to manage a zero-carbon electricity system, and must develop new tools and markets to replace the traditional steam and gas turbines it once relied upon.  Put together, these not only change the size of failures that could be expected, but also the location of them, and the balance between centralised and distributed sources (the latter which could potentially be invisible or uncontrollable to the system operator).

Demand will also change dramatically, as efficiency improvements continue to shrink the traditional pattern of consumption, and new technologies like electric vehicles and heat pumps come into the mainstream. Smart systems mean consumers could be a major part of the solution, with power companies quick to offer ‘smart hot water tanks’ that will help in the event of blackouts in future.

While Britain has one of the most secure electricity systems in the world, it is worrying that events like this expose several flaws.  It is little comfort to those affected to say the power system worked exactly as expected. Engineers will continue analysing the events of August 9^th for years to come, but they must also be quick to develop new ways of keeping Britain’s power system under control as it moves into bold new territory.

The lead author of this special report is Professor Tim Green from Imperial College London. Dr Iain Staffell is main author of the Electric Insights Quarterly report in which it first appeared.

Electric Insights is commissioned by Drax and delivered by a team of independent academics from Imperial College London, facilitated by the college’s consultancy company – Imperial Consultants. The quarterly report analyses raw data made publicly available by National Grid and Elexon, which run the electricity and balancing market respectively, and Sheffield Solar. Read the full Q3 2019 Electric Insights report or download the PDF version.