1. Gemma Kelly, Alan Thomson
Evaluating the use of geomagnetic indices for identifying potential damage to power grids
Gemma Kelly, Alan Thomson

2. GICs
Geomagnetically induced currents (GICs) are the main space weather risk to power grids
We would like a way to categorise GICs, particularly for the UK, to better identify times when the system is at risk.
Ideally we could then provide a forecast with a focus on GICs
We generally assume GIC is most closely related to dB/dt – how true is this?

3. GIC data
Not particularly extensive – relies on a close relationship with Scottish power – they don’t save the data so we have to remember to ask for it in a timely manner
Max measured ~40A

4. dH/dt and GIC
So first assumption look like it holds, GIC follows the dB/dt pretty well. Unfortunately……
March 2012 – Torness GIC data (red) against dH/dt at Eskdalemuir (blue)

5. dH/dt and known GIC impacts
Storms in red are identified in Erinmez – or in the case of 2003, engineers at the time reported some effect (noise, heating etc)
Erinmez et al., J. Atmos. Sol-Terr. Phys. 2002

6. Problem Our earlier GIC data need reprocessing
dB/dt at individual stations is very hard to predict (at the moment)
We would like to develop an index to identify when there is a risk of large GIC
We start by investigating what existing indices GIC (dB/dt) in the UK relates to.

7. NOAA G scales and Kp Widely used and recognised
We use them to forecast and to categorise past activity
Based on Kp, the 3-hourly geomagnetic index
Commonly used – supply forecasts and post-event analysis in terms of Kp – how useful is that for GIC? NG tell us they aren’t concerned until well into G5 – how good an assumption is that? Is that sensible?

8. dH/dt and Kp
So Kp 8 predicts events in the red zone – but also a lot of times when dH/dt didn’t get larger than 200 in that period
At the moment with our current data we don’t know whether any of the other Kp 8 or 9 events caused lesser effects in the power grid (maybe later on or degradation rather than instant effects)

9. Max dH/dt at 24 observatories
dH/dt and Kp
24 observatories with >16 years data
Largest dH/dt at Kp = 9o for only 8 of 24 observatories.
dH/dt > 1000nT/min was measured at 4 observatories: BRW, ABK, ESK and FCC (all > ±57° geomag latitude)
Max dH/dt at 24 observatories
Selection of global observatories with at least 16 years of data
Important to note that the largest variations in X and Y do not always occur on the same dates.
Unsurprisingly a mid-latitude measure of activity not great for near poles

10. dH/dt and K
Regional K better – in that it at least identifies all events as K=9….but still a large range of dH/dt at K=9

11. dH/dt and AE AU AE AO AL 1 minute AE against 1minute dB/dt
AL is westward electrojet
AU eastward

12. dH/dt and AL
Interesting that we only see dH/dt > 600 when AL is small – is this a sign that the auroral oval has moved too far south of the AE stations?

13. dH/dt HSD
Hourly standard deviation – rolling for last hour

14. So how does this help?
We can try using some thresholds to see if that helps narrow down times of increased GIC, for example:
Kp > 8
AL > -1000
HSD > 100

15. Selected data We identify all the storms with dH/dt > 600nT/min
Fewer times identified where there is no large dB/dt at some point during the storm

16. Future work
= GIC measuring location
Reprocess the GIC data we hold and repeat and expand the process with that
do the other indices add value to dB/dt?
Build regression relationships between GIC and other variables
Better proxy than ‘just’ dB/dt
Forecasts and validation?
Do something similar with NZ data?
Comparative study, similar latitudes
Investigate whether the dB/dt at Esk is best for UK GIC or whether some combination of data from UK observatories is better

19. Quiet
High GIC risk
Medium risk

20. Very simple approach => large number of false alarms
All data Points
Accuracy (fraction correct) = correct/total
= ( )/ =
Predicted
Observed
No storm
Storm
7433
storm 8
False alarm ratio = false alarms/(hits + false alarms)
7433/( ) =
Max per day
Predicted
Observed
No storm
Storm
10888 66
storm 4