Recent advances and validation of GIC modelling in the UK Gemma Richardson, Ciaran Beggan, Alan Thomson

Modelling GIC Models of geomagnetically induced currents (GIC) require: Input geomagnetic field Electric field calculation (including ground conductivity model) Estimation of GICs in the network We have recently been upgrading our network and trying to validate each step in the process Look at the three things above and the full process

Input geomagnetic field How does distance from observatory affects GIC estimates? Use Spherical Elementary Current Systems (SECS) to interpolate magnetic field Compare with using data from individual observatories across the whole grid

Input geomagnetic field Results shown for October 2003 storm Observatories within the network, or on the same geomagnetic latitude within 400km gave estimates within 25% on average with correlation >0.8 More distant observatories become less reliable, for example experiencing max GIC at different times

Electric field calculation Measurements We have electric field measurements at the three UK observatories 10Hz Probes ~100m apart in NS and EW directions Conductivity model using data from airborne geophysical surveys of UK and associating with geology from the 86 lithologies using UK geological map And 10km grid So we will obviously not have the local detail that will be included in the measurements – but during a storm when regional signal is larger we would hopefully get a bit of a match

Electric field calculation Model Thin-sheet model The conductivity model of top 3km based on UK lithologies and bathymetry 10km grid Anomalous conductance

Electric field Measured (blue) and modelled (red) E at 1 minute resolution With a 10minute moving average on the measured data to smooth it (used 10minutes for E calculation) Not included Lerwick as its so far off the model anyway Hartland still very affected by tidal signal – will look at removing that this week Overestimate hartland Y particularly – still need to to a lot of work on understanding local effects Correlation NS : 0.43 EW : 0.43 Correlation NS : 0.04 EW : 0.29

Electric field We need to account for local effects better, e.g. galvanic distortion, tidal signals, local conductivity etc. (see poster 5 – Baillie et al.) More storms will help During the SWIGS project (see Poster 4 – Thomson et al.) we will measure E-fields at more sites in the UK so we can also test the model away from the observatories

GIC calculation Horton et al 2012 (IEEE Transactions on power delivery, 27) provides a test grid and calculated GIC for a uniform electric field Mix of transformer types, single and parallel connections and blocking devices

Comparison Despite some differences in the method we get results which are consistent with the Horton paper Horton et al. (2012) BGS North East Sub1 0.00 Sub2 115.63 -189.29 114.25 -189.77 Sub3 139.85 -109.49 137.87 -109.79 Sub4 19.98 -124.58 19.22 -124.63 Sub5 -279.08 -65.46 -280.55 -63.94 Sub6 -57.29 354.52 -53.24 353.99 Sub7 Sub8 60.90 134.30 62.45 134.14 We use a more sophisticated approach for getting the voltage - they just do V = E_N*L_N + E_E*L_E - which is fine here as E is constant. We use nodal admittance method - The root-mean-square differences is 2.3 A for the North direction and 0.7 A for the East direction Per transformer we the difference was <1A per phase for all but 3 transformers and they were all <2A

Power network upgrade Network upgrade using data from the 2016 Electricity Ten Year Statement Better representation of substation nodes Better inclusion of parallel lines

Comparsion with GIC measurements Correlation 0.469 0.448 0.503 0.615 Quite an underestimate at Strathaven – which is odd – with the EURISGIC grid Strathaven was pretty good I think. I tried changing the way the grounding is done at this one site (as it was one with separate earthing mats) but it didn’t make much difference. I’m not sure whether changing that for all the sites will change anything Maybe Strathaven is related to the e-field underestimating Esk? Hartland E isn’t great but its far from these particular sites

Summary Electric field on a broad scale we capture the storm need to account for local effects better to truly compare the model and measured data Future work will help with validation across the UK Validated GIC calculation Full GIC calculation Getting the size and timing of GIC largely correct although these are relatively small GIC measurements SWIGS project will really help with validation – providing more measurements of both E and GIC

Ler – Mar 2015

Had - unfiltered 0.1406 0.0314

Sep 2017 – E-field

Sep 2017 - GIC correlation HUNT 0.5343286546782464 correlation NEIL 0.1842394052155707 correlation STRA 0.438831398654849 correlation TORN 0.4891124941799598

Mar2015 resampled to 1min means