There has been lots of discussion on what caused the nationwide power cuts we saw last week and what the implications are for our industry. Most of this discussion is highly technical, and while for engineers is exciting, there are many around the industry who want to know what happened, but are not so interested in what ROCOF, LFDD, FFR, EFR all mean!
So this is my version of what happened, hopefully designed for non-highly technical engineers to follow.
First some power system basics. The GB electricity transmission network, which was designed to connect large generators to all the local lower voltage Distribution Network Operators (DNOs) is a completely synchronised system. This means that it is all interconnected, with the same frequency wherever you are in GB.
The Frequency of the grid is very important because certain devices need a fixed frequency to operate correctly (for example clocks). Frequency is related to the balance of generation and load on the system – if it balanced perfectly, the frequency is 50 Hz. If there is more generation than load, the frequency will rise and if there is more load than generation, the frequency will fall.
National Grid Electricity System Operator (NGESO) has the job of managing this balancing act to ensure frequency stays within statutory limits. Under normal conditions the frequency must be maintained between 49.8 Hz and 50.2 Hz. Under exceptional events the frequency should not go outside the range 49.5 to 50.5 Hz for more than 60 seconds.
NGESO manage this by ensuring generation always matches demand. This is a difficult task – most people are aware of ‘TV Pick Ups’ for example, when a popular TV programme ends and lots of people boil a kettle, adding instant massive demand to the network. This is all planned in advance with fast acting generation arranged to kick in to ensure balance is achieved.
The hardest challenge however is when something causes a large generator to ‘trip’, removing a large infeed from the network. This is not that uncommon – it could be a high voltage fault or a fault within the generation plant. NGESO plan for the ‘largest infeed loss’, which is currently 1,400 MW. To manage a sudden loss of this scale, NGESO procure ‘ancillary services’ from generators or other providers, who can react very fast to an instruction from NGESO to increase generation to regain balance.
If the frequency is ever allowed to fall too far, all the generation will ‘trip off’, and the entire GB national grid will go dead. This has never happened and would cause massive disruption to the country. For this reason, there are emergency automatic systems in place to prevent this happening, these systems did act last week, as they were designed.
Through the Grid Code, which all DNOs must comply with, they are required to have in place a scheme known as ‘Low Frequency Demand Disconnection’. This is the last line of defence in a major system frequency excursion – when the frequency hits certain thresholds, relays within the DNO substations will ‘trip’ the substations off the network, shedding loads in order to save the wider system. This progressively happens ultimately shedding 60% of the DNO loads. Of course, this DNO load is actually houses, schools, National Rail, hospitals etc.
In the past, most of the generation in GB was made up of large power stations, with steam driven turbines rotating at 3,000 rpm, giving the system its frequency of 50 Hz. The massive spinning turbines are extremely heavy, so when generation was lost, and the frequency started to drop, these large turbines resisted this with their high mechanical inertia, ‘ridding through’ the event and maintaining the frequency until more fast-acting generation was brought on-line.
These days, the number of large turbines connected to the grid is less. We have more interconnectors with other countries supplying power, and much more wind and solar generation. None of these types of generation provide any inertia to the system. This is where battery storage comes in – batteries can act very fast to an event and export large amounts of power to quickly support the grid to restore frequency to the correct range. This is one of the ‘ancillary services’ NGESO procure.
We don’t have the full details yet, but this is my view on what happened last Friday, all illustrated on the graph below.
At 16:52, Little Barford, a gas fired power station near St Neots in Cambridgeshire tripped, taking 664 MW of generation off the system instantly. Within 1 second, the frequency dropped to 49.3 Hz, but started to recover, likely as a result of NGESO ancillary services kicking in (automatic response of generators and battery storage instantly exporting under Fast Frequency Response contracts). The frequency was well below statutory limits, but was recovering. However, at 16:53, Hornsea 1 windfarm in the North Sea suffered a trip, taking a further 756 MW of generation offline, taking the total missing generation to 1.4 GW (which is also the ‘maximum infeed’ level that NGESO plan for).
This second trip caused a further fall in frequency, which overwhelmed the ancillary services being utilised to stabilise the first frequency drop. The frequency quickly fell to 48.8 Hz, low enough to trip the first stage of the DNOs Low Frequency Disconnection relays, shedding load all over the country (roughly 1.2 GW was shed by the DNOs, but this is an approximation as this is not clear yet).
This shedding of load combined with the ancillary services that NGESO were bringing to bear recovered the frequency to stability by about 17:06. At that stage, with further generation now contributing to the system, NGESO was working with the DNO control rooms around the country to restore the substations that had been tripped off, restoring electricity availability to everyone.
What took much, much longer seems to have been restoring the transport networks – with signalling and stalled electric trains being the main issue. As the evening went on, this then turned into a major logistical challenge – trains stranded, crews in the wrong place and working hour limits being potentially breached. This disruption lasted long after all supplies were restored.
Just to make things more complicated, we don’t fully understand yet what happened to the ‘embedded generation’. This is smaller scale generation that is connected to the DNO networks, so effectively invisible to NGESO. This generation is not allowed to ‘island’ operate, so if the network it is connected to goes dead, it must trip itself off, using parameters laid down by the industry ‘G59’ standard. Unfortunately, how these G59 relays detect the network connection failing is by using a method called ‘rate of change of frequency’ (ROCOF). A large frequency excursion can trip these relays (as happened in the last large frequency excursion in 2008), removing further generation from the network, making the generation imbalance even worse. An attempt was made since 2008 to change the settings in these relays, it will be interesting to see what happened in this case.
One thing we should be clear on – something did go wrong, but we did avoid a complete system shut down, and a subsequent black start. That is a much, much worse scenario. There is some modelling that suggests it may take 72 hours to restore the system in a black start scenario.
There are a number of implications and questions from the above, which will need real data and technical understanding to answer:
Why did the loss of 1.4 GW, which is within the planned maximum in feed loss criteria, cause such a large frequency excursion?
Inertia, as measured by NGESO was within system parameters, are these parameters correct?
Was the second trip somehow connected to the first trip (ROCOF?)
What role did embedded generation play?
Was enough operating reserve procured by NGSO?
How much contribution did FFR battery storage have?
Do we need to re-think Low Frequency Demand Disconnection, and in particular what exactly it trips?
Why was there such a massive impact on the transport industry, in particular trains?