Why did the lights go out?

On Friday, 9 August, there was a series of power-cuts across England and Wales. Swiftly, Ofgem, the Government regulator for the electricity market, announced an investigation into what happened and National Grid ESO have now published an interim technical report. The full report is due in early September, but the interim report tells us a lot. Professor Tim Green has taken a look at what we now know about what happened that day.

The report gives us a very interesting account for engineers looking to understand what happened on that day almost two weeks ago. The sequence is more complex, and subtle, than was evident on the day itself.

The initial cause was a lightning strike. Lightning affects power lines many times a year and is routinely dealt with. This one struck near, or on, the overhead line connecting two substations at (Eaton Socon near St Neots and at Wymondley near Hitchin). Large currents were seen in the substations, and circuit breakers opened to disconnect one of the two lines on that route (in this case it only took 70 milliseconds). Those circuit breakers re-closed automatically, after about 20 seconds in this case, and stayed closed because the “fault” had gone, that is, the large current caused by the lightning strike had ceased. They would have re-opened and locked open if a permanent fault such as a damaged line had been the cause.

And hundreds, perhaps thousands, of times year that is the end of the story. There are parallel circuits to the one that disconnects, present as part of the grid’s built-in redundancy, and no generators or customers are affected.

But it was different that day …

The disturbance to the voltage caused by the lightning and the line outage then appears to have led to three things to occurring at essentially the same time (within half a second of each other).  First Hornsea offshore wind farm saw the voltage changes and attempted to help correct them. But then an internal anomaly was detected, which caused two modules to abruptly disconnect taking 737 MW of that wind generation off the grid.

The second issue was that the lightning strike disturbed the Little Barford power station; it is connected to Eaton Socon substation at one end of the affected line. It is a combined cycle plant, with two gas-turbine generators and a steam-turbine generator (the steam is raised from the heat in the exhausts of the two gas-turbines). The disturbance lead to anomalous speed measurement data at the steam turbine and so it automatically shut-down taking an additional 244 MW offline.

This voltage disturbance rippled out across the system resulting in sudden shift in angle of the voltage (the starting point of the AC voltage cycle measured against a reference point). That matters because “distributed” generators, like small wind and solar plant, use this as one way to detect “loss of mains” and about 500 MW of these small generators detected this as problem and disconnected creating the third loss. Now with a total loss of about 1,480 MW of generation, the frequency started dropping as remaining generators slowed down while still supporting 33,500 MW of demand. It dropped quickly to 49.1 Hz.

However while it was dropping the first line of defence, the cavalry if you like, was coming to the rescue in the form of 650 MW of “response” power. That was a mixture of part-loaded generators, batteries and demand response (voluntary reduction of load from some customers, mostly commercial and industrial customers) that National Grid Electricity System Operator (NGESO) had contracted to be in place. The batteries will have been very fast and undoubtedly helped a lot.

This was not the only backup and within 20 seconds a further 350 MW of response arrives. This brings the total backup deployed to 1,000 MW which is what NGESO was holding to cover an outage of the largest generator running that day (sometimes it holds more, 1,300 MW if bigger units are running). The mix of response is very interesting too, 200 MW of generation; 450 MW batteries and 350 MW demand response. Only 10 years ago it would have been very different, it would all have been generation (part-loaded gas or coal fuelled generators).

So far the system was working as it should, no demand disconnection, only contracted (voluntary) demand response. The system was still short of generation but a 480 MW gap is not insurmountable, the frequency had stabilised and further actions, secondary reserve would be called up to restore the frequency to its normal operating range of 49.2 – 50.2 Hz. But then there was some bad news and a series of unfortunate events.

With the steam unit shut down, the gas units at Little Barford can’t operate. The first gas unit tripped about 1 minute after the original event losing a further 244 MW. We’re now down 1,691 MW with a reserve of 1,000 MW already deployed.

At this point frequency falls again, now to 48.8 Hz, which is where the second line of defence is pre-programmed to come in. It’s called low-frequency demand disconnect (LFDD). This disconnects about 5% of demand, it was 1.1 million customers on this day. More would have been disconnected at 48.7 Hz but fortunately frequency started to recover.

Over the next 4 minutes NGESO control room called up 1,240 MW of further action (secondary reserve) and got the frequency back to 50 Hz. During this process the second gas unit at Little Barford tripped too but the effect was minor. After a further ten minutes, when NGESO were confident the recovery was secure, they instructed distribution network operators to start reconnecting customers.

Timeline of events (from the Interim Report)

I think it is worth noting that the time between the lightning strike and starting to reconnect consumers was 15 minutes and all customers were reconnected 35 minutes after that. The detailed reasons why the railways were so badly affected by these events is outside the remit of this report but is an issue that needs investigating. The report does say that the railway systems were not generally part of the LFDD disconnection but they reacted to the perturbation of the grid frequency and voltage. This did cause serious and longer-term impacts on the public and lessons will need to be learnt.

So, what do we make of all this?

  1. It is very rare to see large power stations disconnected in response to lightning strike and/or line outage. Losing two to the same event is exceptionally rare and in that light the system responded well.
  2. The last time two generators were lost in quick succession was May 2008 with half a million customers lost to LFDD, so we can say it is rare.
  3. LFDD is very painful for those affected but 5% suffered so that 95% stayed connected to a functioning grid.
  4. There will be technical lessons learned by the owners of Little Barford power station and Hornsea wind farm and there will perhaps be implications for other generators.
  5. The rate at which frequency changed, 0.16 Hz/s, was high in historic terms but to be expected with our new low-inertia grid (wind turbines and solar panel don’t naturally contribute to the spinning mass of the generator fleet so frequency moves faster when supply and demand are out of balance).
  6. The mix of response was healthy – the contracted batteries and demand-response were quick to act and that is why they are used in low-inertia grids.
  7. Could this be prevented? If NGESO had been holding 2,000 MW of reserve, not 1,000 MW, then no LFDD would have been needed. NGESO spends close to £300M/year contracting response/reserve. Spread across 25 million customers that’s a £12/year component of your bill. Should we double it? Perhaps we should have a public discussion of whether we want to protect ourselves from outages of 2 large generators, or 3 or 4 and what we would be prepared to see in our bills to have that provision.
  8. Could more be done to refine “loss of mains” protection to avoid losing distributed generation? Actually this is already underway anyway. There is lots of technical information available online.
  9. Is wind a problem? Well first off, UK has halved its carbon emissions from electricity from 150 million tonnes per year in 2012 to under 70 last year and wind was a big part of that. But yes, there are challenges to address as we add more wind and decarbonise more deeply. We need to address further reduction in inertia (seen over seconds), covering short-term fluctuation and forecast errors (over an hour), periods of low wind speed (days) and seasonal variations (more demand on dark winter evenings). These challenges are acknowledged across the industry and are being addressed.

Our grid system is part way through a major transition because it is in the vanguard of reducing to zero the carbon emissions of our whole country. Clearly, that transition must be done in way that means that power cuts are very rare events. This is why I, and thousands of other engineers, are in the game; we love solving tricky problems that have important benefits to society. This is what we’re all talking about.

The full Ofgem interim report is available online as a PDF.

Professor Tim Green

Tim Green is co-director of Energy Future Lab, Imperial’s cross-disciplinary energy institute with responsibility for facilitating interdisciplinary research and postgraduate education in energy and in coordinating its dissemination. He is also Deputy Head of the Department of Electrical and Electronic Engineering.

His personal research focuses on the analysis and technology to support the development of a low carbon electricity supply network that is able to accommodate variable renewable sources and new widespread electric vehicle charging while still delivering a cost effective and very reliable service. His research specialisation is in power electronics for use in power systems.