Validating modelled transformer-level GIC flow in New Zealand’s South Island with extensive observations
3 year MBIE funded project Solar Tsunamis: Mitigating Emerging Risks to New Zealand's Electrical Network
Photo: Benmore HVDC main AC side transformer
Tim Divett1, Craig J Rodger1, Daniel Mac Manus1, Malcolm Ingham2, Michael Dalzell3, Ciaran Beggan4, Gemma Richardson4, Ellen Clarke4, Yuki Obana5, and Alan W P Thomson4
1. Department of Physics, University of Otago, Dunedin, NZ
2. School of Chemical and Physical Sciences, Victoria University of Wellington, NZ
3. Transpower New Zealand Limited
4. British Geological Survey, Edinburgh, United Kingdom
5. Osaka Electro-Communication University, Osaka, Japan
A large geomagnetic storm started on 6 November 2001 at ~2:53pm LT (1:53am UT). At this time HWB T4 (Dunedin) tripped, as did systems at ISL (Christchurch). Alarms occurred at multiple locations across the South Island, including OHA, OHB, and CYD.
The transformer at Dunedin / Halfway Bush (HWB T4) suffered a major internal flashover. A subsequent internal inspection found the transformer was beyond repair - it was subsequently written off (~$2 million value in 2016 NZD).
Effects of GIC on power systems - NZ
HWB T4 (written off)
ISL SVC (tripped)
= Alarms
Extensive GIC observationsTranspower NZ Ltd measures and archives transformer neutral currents at - 35 transformers in 2001- 64 transformers in 2015In many substations currents are measured in several transformers
Mac Manus et al., Space Weather, 10.1002/2017SW001635
Rodger et al, Space Weather, 10.1002/2017SW001691
Method: Modelling in New Zealand’s South Island
Method: Modelling in New Zealand’s South Island
Method: Modelling in New Zealand’s South Island
Divett et al., Space Weather, 10.1002/2017SW001697
Method: Modelling in New Zealand’s South Island
Divett et al., Space Weather, 10.1029/2018SW001814
Divett et al., Space Weather, 10.1002/2017SW001697
Observations St Patrick’s Day Storm 2015
Representative storm
We had 41 transformers monitored for GIC – 27 with reliable measurements
Observed ~50A at HWBT4 at sudden commencement
Observed several peaks up to 20A for ~30 min for next 14 hours
+ EYR
. HWB
1st attempt at Validation
Try what Beggan et al. did for the UK
- input time series of B field
- linear fit to geomagnetic latitude (not SECS b/c not enough east-west points)
- calculate E field for T = 10 min
- calculate timeseries of GIC
- compare with 10 min mean of observations as a low pass filter of observations
- looks more like B than GIC which looks more like dB/dt
2nd attempt at ValidationTry what Bailey et al. (2018) did for Austria
- Time series of dB/dt field
- linear fit of dB/dt to geomagnetic latitude
- calculate E for T = 10 min
- compare with 10 min mean of observations
- multiply modelled GIC by 500 to get on same scale
- sudden commencement about right
- doesn’t reproduce any of the later GIC peaks at all
3rd attempt at Validation - do it in the frequency domain
Time domain Frequency domain
Single pointB observations
Spatially varying B field
Spatially varying E field
GIC
FFT
Linear fitamplitude and phase
Thin sheet model
Network modelFilter
& IFFT
3rd attempt at Validation - do it in the frequency domain
Time domain Frequency domain
Single pointB observations
Spatially varying B field
Spatially varying E field
GIC
FFT
Linear fitamplitude and phase
Thin sheet model
Network modelFilter
& IFFT
Validation in frequency domainMagnetic field 27/3/2015
+ EYR+ TEW
+ MDM
+ MCQ
Results: Efield for period of 23.6 min
Validation: GIC in frequency domain
Validation: GIC in frequency and time domain
Validation: GIC converted to the time domain
Sudden commencement not as high as raw but similar to filtered observations
Later peaks match timing of filtered observations
Amplitude of later peaks not always perfect but not too bad
Validation: GIC in frequency domain19/27 ‘Good’ matches (70%)
8/27 ‘Bad’ matches
Exclude transformers where <1A measured
18/21 ‘Good’ matches (86%)3/21 ‘Bad’ matches HWBT4L, ISLT6H, OHAT7H
Validation: GIC in frequency domain19/27 ‘Good’ matches (70%)
8/27 ‘Bad’ matches
Exclude transformers where <1A measured
18/21 ‘Good’ matches (86%)3/21 ‘Bad’ matches HWBT4L, ISLT6H, OHAT7H
Comparing spatial variation with 2001 storm events
SDN
TWI
KIK
ISL
HWB
}
}}
HWB T4 (written off)
ISL SVC (tripped)
= Alarms
+KIK
+TWI
+ SDN
+TKB+OHBOHA +
} OHBOHA
TKB
single period = 26.3 min
Conclusions and Future Work
Adapt BGS’s Thin Sheet and Network Model’s to NZ
Build conductance model Modify network model to calculate
transformer-level GIC Validate model with Transpower’s
GIC data in frequency domain Validate model in time domain
- Range of frequencies we can simulate is limited at high and low T- Transformer-level detail is very important - KIK may be more affected by GIC than realised in the past
- The transformers with by far the highest observed GIC are modelled worst HWBT4 and ISLT6 modelled GIC is lower than measured
- Better understanding of spatial variation of B field would improve modelling
- More detail of conductance, particularly around Otago, might improve modelling- Seems that the thin sheet model should be used in the frequency domain