Geomagnetically induced
currents
Antti Pulkkinen
NASA Goddard Space Flight Center
(antti.a.pulkkinen@nasa.gov)
1
Contents
•  Motivation/background – why should we care?
•  Some observational facts of interest about
geomagnetically induced currents (GIC).
•  Theory and modeling.
•  Forecasting – the Solar Shield project.
•  Future challenges.
•  Summary.
2
Geomagnetically induced current (GIC)
3
The condition of twelve 400 kV GSU transformers, each
rated 700 MVA, at the Tutuka and Matimba power stations
and six 275 kV GSU transformers at Lethabo power station is
checked regularly, with some units equipped with on-line
DGA instruments. After the severe geomagnetic storm at the
beginning of November 2003, often referred to as the
“Halloween storm”, the levels of some dissolved gasses in the
transformers increased rapidly. A transformer at Lethabo
power station tripped on protection on 17 November. There
was a further severe storm on 20 November. On 23
November the Matimba #3 transformer tripped on protection
and on 19 January 2004 one of the transformers at Tutuka was
taken out of service. Two more transformers at Matimba
power station (#5 and #6) had to be removed from service
with high levels of DGA in June 2004. A second transformer
at Lethabo power station tripped on Buchholz protection in
November 2004.
The DGA records are not the same for all the transformers,
but all of them show a sharp change at the end of October
2003, when the first storm occurred. Based on the DGA
records, most of the transformers at these power stations
appear to have been damaged by the effects of the
Typically, on four apparently damaged transformers:
Ethylene:methane C2H4 / CH4 = 0,2 - 0,9
Ethane:methane C2H6 / CH4 = 0,2 - 1
Methane:hydrogen CH4 / H2 = 2 - 5
Ethylene:ethane C2H4 / C2H6 = 0.4 - 4.6
Acetylene C2H2 negligible
In the transformer depicted in Fig 6 the level of CO2, a
product of low temperature degradation of cellulose, was
approximately 10 times higher than the level of CO.
Relatively higher CO or ethylene content would indicate
higher temperature degradation.
Inspections of all the failed transformers identified heat
damage, mostly to paper insulation, in various parts of the
transformers, as illustrated in Figs 6 to 8. The damage is
consistent with the DGA results. In all cases, the extent of the
damage appears to be small, and discoloration of paper
insulation beyond the immediate vicinity of the fault is
superficial, which explains why the absolute levels of
dissolved gas are low - even below the threshold considered
significant for most DGA assessment.
Why do we care? (applications)
• GIC causes saturation of power
transformers:
2012 Special Reliability Assessment
Interim Report:
Effects of Geomagnetic
Disturbances on the Bulk
Power System
• Transformer damage
February 2012
• Electric blackout
Fig 6: Failure in HV winding of Lethabo #6
Fig 7: Failure in HV winding of Matimba #4
Fig 8: Overheating of LV terminals of Tutuka #1
4
Why do we care? (physics)
5
What
does
GIC
look
like?
Typical amplitudes
GIC observed on October 29, 2003
100
80
Bursts
60
Sudden start
40
GIC [A]
20
0
−20
−40
−60
−80
−100
Complex waveform
06:00
09:00
12:00
15:00
UT [hh:mm]
18:00
21:00
6
57° geomag.
lat.
What drives (large) GIC?
HUTTUNEN ET AL.: SOLAR WIND DRIVERS AND GI
S10002
Table 1. Statistics of the 60 Large Amplitude GIC Events
During the Solar Cycle 23a
N
hjGICjmaxi (A)
N10
hjEYjmaxi (mV/m)
hjdBZ/dtjmaxi (nT/s)
h(Pdyn)maxi (nPa)
h(!)maxi (1013 J)
h(W!)maxi (1016 J)
Sheath
Ejecta
BL
CIR
Shock
13
24.2
5
19.3
0.80
27.5
1.4
3.6
9
17.5
1
17.8
0.37
12.1
1.0
3.0
10
22.4
4
24.1
0.82
22.5
1.8
3.4
2
19.8
0
7.2
0.38
7.1
0.19
0.36
1
11.4
0
a
Rows from top to bottom the number of GIC > 10 A in each driver
category (N), the averages of the maximum jGICj, the distributions of
the 20 and 10 largest GIC between different driver categories (N20 and
N10), and the last five rows give the averages of the maximum values
of the selected solar wind parameters (see text for details). Only the
events that occurred within the local midnight of the Mäntsälä station
(2000 -- 0400 LT) have been considered in this study.
close to !100 nT or belo
ated GIC with BZ only
remained the same even
increased to 3 h.
[21] Correlations betw
Dst and Kp indices are
Correlation is significan
plitude and Kp than bet
with correlation coeffici
Only ejecta associated ev
between GIC and Dst. In
that all large GIC valu
values (i.e., Kp > 5), but
no significant geomagne
of Dst. This is consisten
[2006a] who found that
the local K index exceed
4. Discussion and C
Huttunen et al. (Space Weather, 2008)
scatters in the relationships. It is also seen that for sheath
and BL categories the range of solar wind parameters is
greater than for the ejecta category and it appears that the
[22] In 7this paper we h
the largest GIC (amplitu
part of the Finnish pipel
Solar wind drivers wer
# of 60 s values
5
10
4
10
3
10
2
10
1
10
35° geomag.
10
15
lat.
0
magnetic field observations carried out at the Memanbetsu
geomagnetic observatory. As shown by the validation of
the model in Section 2, this goal was accomplished and
the model can be used for further analysis of GIC in Hokkaido, Japan. Consequently, it is of interest to move
beyond the plain model construction and use the model
features in further interpretations, which are discussed in
this section.
In an ideal case of true 1D ground conductivity structure the computed system parameters (a, b) contain information mainly about the electrical and the topological
characteristics of the power transmission system under
investigation. However, the geological and geoelectric situation in Hokkaido, Japan is complex and 3D (see, e.g.,
Satoh et al., 2001; Ichiki et al., 2009). Consequently, one
What drives (large) GIC?
10
5
0
5
GIC [A]
10
15
Fig. 6. Statistical occurrence of modeled GIC in Hokkaido, Japan. Data
for years 1986–2008 were used in the construction of the statistics.
Table 1
The list of 10 largest GIC events between 1986 and 2008. The maximum absolute value of the modeled GIC was used to measure the strength of an event.
The last column indicates the geomagnetic storm phase associated with a modeled GIC peak value.
Year
Month
Day
Hour
Minute
jGICj [A]
Storm phase (CME-driven)
1989
2003
1986
2004
2001
1999
1991
2000
1991
2004
03
10
02
11
11
09
03
07
06
07
13
30
08
10
06
22
24
15
13
27
22
19
20
09
02
21
04
21
01
11
09
55
51
42
03
51
03
37
11
57
25
21
17
15
15
14
13
12
12
12
Early main phase
Early main phase
Early main phase
Late main phase
Early main phase
Late main phase
Sudden commencement
Late main phase
Sudden commencement
Late main phase
Please cite this article in press as: Pulkkinen, A., et al. Modeling geomagnetically induced currents in Hokkaido, Japan. J. Adv. Space Res. (2010),
doi:10.1016/j.asr.2010.05.024
Pulkkinen et al. (JASR, 2010)
Coronal mass ejections the most significant driver of
large GIC at mid- and high-latitudes.
8
nt control surveys.
his paper is as follows: We start by
e GIC measurements in the Finnish
(section 2). We also discuss the
magnetic indicators for quantifying
y, we outline the theoretical basis of
e, its practical operation, and its
s of the direct GIC measurement
ents
57° geomag.
lat.
the old site, as deduced from model calculations. However,
the daily GIC range is about 15% larger than earlier on
average. To have a strictly uniform time series, we only
present statistics until spring 2005 in this paper. After
spring 2005, the geomagnetic activity has been very low
because of the approaching sunspot minimum, so we do
not miss any major event until the present.
[10] We characterize the daily GIC activity by two measures in Table 1: the maximum absolute 10 s value and the
number of 10 s values exceeding 5 A. We also show the
local magnetic activity indices K and Ak. The K index is a
What drives (large) GIC?
he nowcasting service is based on
ing along the pipeline at the Mänon since November 1998 (Figure 1).
em resembles the idea used for the
Campbell, 1980]. It consists of two
ust above the pipe and another at
ysical Observatory [Pulkkinen et al.,
t above the pipe records both the
ariation and the field due to GIC
e. The reference magnetometer at
only the natural variation. Since the
storm-time
tsälä and Nurmijärvi
is only about
ariation fields are approximately
s [Jankowski et al.,pulsations
1986]. So the
s GIC by inverting the Biot and
rence magnetometer is not in the
e spatial variation of the geomagsome small errors. Instrumental
ctromagnetic noise, hard weather
ovements due to frost are seen in
storm-time
substorms
Figure 2. Diurnal occurrence of 10 s GIC values
exceeding 2.5 A at Mäntsälä in November 1998 to
May 2005.
Viljanen et al. (Space Weather,
2006)
2 of 9
9
Theory and modeling
•  Modeling of GIC is done in two steps:
–  Geophysical step – compute the geoelectric field using
geomagnetic induction theory.
–  Engineering step – compute the GIC using the known system
parameters.
10
Geomagnetic (quasi-static) induction
in linear, isotropic and nondispersive
medium
Theory and modeling
∇⋅ E =
ρ
ε0
(1)
∂
∇ × E = − B (3)
∂t
€
€
€
j = σE
∇⋅B =0
(2)
∂
∇ × B = µ0 j + µ0ε0 E
∂t
(4)
(5)
€
€
11
€
Theory and modeling
Time derivative of the horizontal
magnetic field
1
Ey = −
πµ0σ
Conductivity
t
∫
−∞
Source field needs to be
known (proportional to
source currents)
gx
dt'
t − t'
History needs to be
known
Conductivity structure needs
to be known
12
€
Complex Image Method
Pirjola and Viljanen (Annales Geophysicae, 1998)
Pulkkinen et al. (PIER, 2009)
Theory and modeling
j (r ')
j (r ')
€
€
σ (z)
ez
€
µ0
€Ai =
4π
Ar =
Ai , Ar
µ0
4π
∫
V
∫
V
j (r ')
dV '
r − r'
C ⋅ j (C ⋅ r ')
(1)
C ⋅j (C ⋅r ')
dV ' (2)
r − r '−2 p(ω,σ (z))ez
€
13
Theory and modeling - GIC
• General solutions for discretely and
continuously grounded systems by Lehtinen
and Pirjola, Ann. Geophys. (1985) and
Pulkkinen et al., J. Appl. Geophys. (2001).
• In practice, GIC can be computed from:
GICi = ai E x (ri ) + bi E y (ri )
System parameters can be solved
empirically or theoretically
Output of the geomagnetic
induction computations
€
14
Forecasting – Solar Shield
•  In Solar Shield, we developed a two-level
experimental system to forecast space weather effects
on the North American power grid; project funded by
NASA’s Applied Sciences Program.
•  NASA/GSFC/CCMC and Electric Power Research
Institute (EPRI) the key players.
•  Pulkkinen et al. (Natural Hazards, 2009).
15
Level 1 forecasts
16
Solar observations of eruptive events
are used to compute “cone model”
parameters. NASA/ESA SOHO/
LASCO
data
used.at the Earth used in a statistical model
MHD
output
Level 1 forecasts
providing probabilistic estimate for GIC at individual
nodes of the power grid. GIC forecast file is generated.
Plasma “cone” introduced to the inner boundary of a
heliospheric MHD model. Model propagates the
disturbance to the Earth. Computations carried out at the
Community Coordinated Modeling Center.
April 3, 2008
17
Level 2 forecasts
18
Lagrange 1 observations used as
boundary conditions for
magnetospheric MHD. NASA’s ACE
Magnetospheric
MHD output used to drive geomagnetic
data used.
induction and GIC code providing GIC at individual nodes
of the power grid. GIC forecast file is generated.
Magnetospheric MHD model used to model the
magnetospheric-ionospheric dynamics. Computations
carried out at the Community Coordinated Modeling
Center.
April 3, 2008
19
Level 2 forecast example
Modeled GIC at lat: 53.2, long: !99.3
10
GIC [A]
5
0
!5
!10
50
100
UT [h]
150
200
Measured GIC at lat: 53.2, long: !99.3
10
GIC [A]
5
0
!5
!10
50
100
UT [h]
Oct 24, 2003
150
200
20
Future challenges
•  Models (coupling, physics etc) and space-based data
drivers (L1, remote solar etc). Utility: forecasting and
modeling of extremes.
•  Magnetometer arrays and corresponding sciencequality GIC observations. Utility: monitoring,
nowcasting, model validation.
•  Geology, especially 3D ground conductivity structure.
Utility: improved modeling of the geoelectric field.
•  Understanding of extremes. Utility: better extreme
scenarios for engineering analyses.
21
Summary
•  GIC global phenomenon that a) impacts long
conductor systems and b) is at the end of chain of
complex physical processes.
•  CMEs the most significant driver of large GIC.
•  Modeling of GIC carried out in two steps - geophysical
and engineering steps.
•  Emerging capability to forecast GIC based on the firstprinciples.
•  Many interesting challenges ahead.
22