9th of August 2019. It’s 4:52pm on a Friday, people are getting ready to go home and enjoy their weekend, then…
5% of Great Britain goes dark.
Wide spread power outages, trains stop, an airport and hospital lose power.
This timeline is based on the interim report from National Grid, published on the 16th of August. It’s a pretty good read! Would recommend. It’s also worth pointing out briefly that all media reporting I’ve seen of the UK blackout have the events wrong – either in incorrect order, or incorrect attribution to causes.
I’ll be focusing on the 76 seconds between 4:52:33PM when the intial event occured, through to 4:53:49 PM when the load shedding occured.
The grid begins in a stable operating state, these next 4 events all happen within 1 second:
Lightning hits the Eaton Socon - Wymondley transmission circuit. A normal and unremarkable occurrence. The circuit disconnects and opens after 70ms to clear the fault. This circuit will re-energise and come back online in 20 seconds. This is good and normal!
The lightning strike created a transient voltage disturbance which caused the loss of 500MW of small embedded distributed generation (solar, small gas and diesel) on the transmission circuit. This is good and normal and meant to happen when lightning strikes a line!
“Hornsea started deloading”. Not good! Hornsea, a large offshore wind farm changes output from 799MW to 62MW, a 737MW reduction in output.
“Little Barford Steam Turbine trips 244MW instantaneously”. Doubly not good!
What begins as a lightning strike cascades to a 1481MW loss in generation.
Frequency begins to fall. Over the next 30 seconds 900MW of backup generation is injected into the system and the frequency drop is arrested at 49.1Hz. This is good!
No blackouts have occurred! The system is recovering! It looks like this will go down as a near miss — an increasingly common occurrence in many grids worldwide. Frequency continues to move back towards the normal operating band without the need for load shedding. Grid frequency reaches 49.2Hz at 45s after the initial disturbance.
An important note at this stage, Little Barford is a combined cycle gas turbine (CCGT) power plant, comprised of two gas turbines and one steam turbine. The heat created by these gas turbines is used to create steam and that steam is sent to the steam turbine. During the lightning strike the steam turbine (STC1) went offline. Without the steam turbine running at Little Barford, steam pressure begins to build up in the steam bypass system of the two operational gas turbines, currently outputting at 397MW.
Little Barford Gas Turbine 1 (GT1A), currently outputting 210MW, disconnects from the grid. This is an automated protection response due to excessive steam pressure in the steam bypass system.
The loss of this 210MW causes the grid frequency to start falling again. Emergency reserves are currently maxed out so a downward frequency trajectory at this point is very bad.
Frequency drops below 48.8Hz which triggers the Low Frequency Demand Disconnection scheme (LFDD). Approximately 1GW (5% of total) of Great Britain electricity demand is turned off.
Little Barford Gas Turbine 2 (GT2A), outputting at 187MW, will be manually disconnected 9 seconds after LFDD begins due to high steam pressure. This increase the amount of load shedding required.
Below is a time series plot of grid frequency from the report (click image to enlarge). You can clearly see here the first fall in frequency, it being arrested at 49.1Hz and starting to recover, then falling again to below 48.8Hz.
Why did the initial generators trip?
Here’s Hornsea’s explanation:
“The equipment at Hornsea saw a system voltage fluctuation with unusual characteristics coincident with the lightning. The initial reaction from Hornsea’s systems was as expected in attempting to accommodate and address the system condition, but very shortly afterwards as the reaction expanded throughout the plant,the protective safety systems activated.”
And Little Barford’s:
“The causes provided by RWE for the initiation of the trip of Little Barford steam turbine (ST1C) was due to discrepancies on three independent safety critical speed measurement signals to the generator control system.. “
We’ll know more from the final report but I’m guessing it’ll be digital electronics miscategorising the situation due to Vector Shifts (aka phase jumps) and/or high RoCoF (Rate of change of frequency) environment.
The system stood up to initial loss of 1481MW. It was the secondary loss of 397MW from losing Little Barford GT1A and GT2A that caused the load shedding. Sure, that initial cascading loss of generation shouldn’t happen, and much of the focus will be on that, but I think some of the focus should be what happened after the initial loss.
The grid remained in an unstable operating state after the initial disturbance, and this is why losing just 210MW from GT1A pushed it over the edge. At the time there was 4000MW of unused generation in reserve. If enough of that reserve generation (gas, hydro) could have increased its output by a few hundred MW total within the 60 second period, hastening the return to the normal operating frequency, the load shedding when GT1A and GT2A tripped may have been avoided.
This event plus some conversations I’ve been having in the last few weeks suggests that while grid operators have some automatic protection systems, much of the operations to get a grid back into a stable operating state when these type of events occur is carried out manually. These manual processes can take minutes (sometimes tens of minutes!). This may have been appropriate when grid generation was comprised of a small number of very large generators, but in modern day grids with an increasing amount of small, distributed generation, you need systems that can identify and correct for destabilising events in seconds.
National Grid will be submitting the final technical report on Friday 6 September.
If you’d like to hear more about this as it develops, and about the energy transition in general, you can follow me on twitter @mitch_oneill or subscribe to my energy mailing list below: