You are on page 1of 4

2007 International Conference on Solid Dielectrics, Winchester, UK, July 8-13, 2007

Application of Finite Element Analysis to externally forced water cooled cable circuit ratings

D J Swaffieldl, P L Lewin' and S Sutton2


'School of Electronics and Computer Science, University of Southampton,
Southampton, SO17 1BJ, United Kingdom
2National Grid, Warwick, Warwickshire, UK
*Email:

Abstract: In order to increase the available current rate and heat capacity in addition to its temperature.
carrying capacity of high voltage cables, utilities may
employ forced water cooling within power transmission Cable losses
networks to remove the heat from cable groups. For Within the cable there are three heat sources, these are;
example the transmission network of England and joule loss, that originating from the resistance to the
Wales includes externally forced cooled circuits, i.e. current in the conductor, the sheath loss, a fraction of
cooled by pumping water through pipes buried in the the joule loss arising from the current flow induced by
vicinity of high voltage transmission cables. At the the magnetic field of the conductor and the dielectric
present time a simple and expedient computer program loss, due to movement of charge within the dielectric.
is used to rate these circuits. This model is similar to The allocation of losses to the cables and the use of a
that of Electra 66. To attribute confidence limits to continuous equation to represent the distribution of
cable ratings attained using this method an extended dielectric loss within the cable is implemented as
2-D heat-transfer FEA model has been constructed to described previously for directly buried cables [1, 2].
allow the calculation of the cable core temperature for a
typical water-cooled three-phase circuit of three single- EXTENDED 2-D FEA METHOD
core cables buried in flat horizontal arrangement within
a backfilled trough. A sensitivity analysis of the model To solve the heat-transfer problem an extended 2-D
to changes in ac resistance, burial depth, dielectric loss, FEA model is constructed. The model works by solving
soil thermal resistivity and surface boundary condition 2-D FEA slices to calculate the temperature distribution
has been performed and is presented. at a given location along the cable length; this uses an
estimate of heat transfer coefficient at the pipe
INTRODUCTION boundary. With this solution of the 2-D section the heat
transferred into the coolant over a section length can be
The results of a study to investigate the conservatism of estimated and the increase in temperature of coolant
different cable rating methods on the current carrying calculated. This allows calculation of a new heat
capability of directly buried cable groups using Finite transfer coefficient for the next slice along. For this
Element Analysis (FEA) and alternative expedient model the 2-D models include solution of the heat-
analytical and Finite Difference (FD) methods has been transfer at the pipe boundary due to forced-convection
presented previously [1, 2]. In this paper the focus is on using water as the coolant. This is simpler in
extending the FEA modeling technique previously construction than a full 3-D model and saves
described to solve the heat-transfer problem for the case computational effort and time to solve, resolution in the
of externally forced cooled cable circuits; where heat is z-plane, along the cable length, must however be
removed from cable groups by pumping coolant through correctly chosen so as not to adversely effect the
pipes laid adjacent to the cable group. This arrangement accuracy of results, investigation of this has been
offers the advantage to the utility of increasing the cable presented previously [4].
circuit current carrying capability and has been
employed most typically where circuits are laid near the The calculation for heat transfer coefficient and
surface or in surface troughs. temperature rise of the coolant along a section length is
performed using the following formulae
To calculate the rating of forced cooled cables is not a
trivial problem since traveling along the cable heat is
given up to the coolant which results in a temperature h= q = Nu.k [W.m-2 IK'] (1)
(TS- Tb) D
profile for the cable and the coolant pipe along the
route. The location of the hottest point is not therefore
known at the outset. In addition to calculate the heat N {8){fReD1
(f 000}Pr 2)
1+{112.7(f /8)1/ (Pr 21)}
N

transferred to the coolant requires knowledge of the heat


transfer coefficient at the pipe boundary, h. This
parameter is dependant on the coolant, its density, flow f = 0.3164Rej025 (3)

1-4244-0750-8/07/$20.00 ©2007 IEEE.


67
GROUND SURFACE
T++ qzDL
Tb,out - Tb,in . (4) -Soil
mcp
A
where the constants and variables are defined in
Table 1.
Due to the interaction between the inner and outer pipes
and cables multiple iterations along the total route 150.0 150.0 150.0 150.0 Cnnductnr
.Dielectric
length are required to achieve convergence of the heat .Sheath
Outer Serving
fluxes, coolant temperatures and cable temperatures
along the route. The convergence criteria used in this Figure 1, Geometric cross-section drawing of forced
study was less than 0.1°C for all cable temperatures cooled cable group, all dimensions in mm.
along the total route length in subsequent iterations; in
practice this always provided convergence better than 2315
a) Water on

0.2°C for the pipe temperatures and convergence was 2300


achieved within 3 iterations. With a solution found the m 2280
load can be adjusted and the model solved iteratively ID 2260
-FEA: 10 degC isothermal
-Isothermal regions method
until the maximum operating temperature limit of the 0 2240
cable is found. 222
222600 800 1000 1200 1400 1600 1800 2000
Burial depth (mm)
Circuit description b) Water off
1450
The cable circuit modeled is a 400 kV three-phase 1400 -FEA:
FEA:
10 degC isothermal, soil
10 degC isothermal, no flow
circuit in a group of three single-core cables buried in FEA: still air convection, soil
flat horizontal arrangement, Figure 1. The group is '51300-
a)
-FEA: still air convection, no flow

directly buried and has four water coolant pipes running -co 1200
near to the cables. Water is pumped in at a temperature
I I

of 10 °C and a flow rate of 4.88 1 s-' per pipe. Water 110600)


800 1000 1200 1400
Burial depth (mm)
1600 1800 2000
flows down the two inner "go" pipes (labeled 2 and 3)
and the two outer "return" pipes (labeled 1 and 4) Figure 2, Rating a) with and b) without water cooling
complete the water loop. Cable dimensions, properties for a range of burial depths.
and model parameters are given in Table 1. This
example has been used to investigate the sensitivity of Figure 2b shows the rating of the FEA models for two
the FEA modeling to material and load parameters. ground surface boundary conditions representing best
and worst case heat transfer conditions, isothermal and
Results still air convection respectively. Also shown is the case
For the case of this single cable group the outer cables, for the pipe with still water removed and replaced with
behave similarly and the central cable has the most soil, effectively the directly buried case. Figure 2b
dramatic temperature gradient along its length with the shows that the choice of ground surface boundary has a
hottest point at the position of the turnaround point for greater influence on the rating at shallower burial
the water pipes. With reference to Figure 2a the rating depths, as the conduction path to the surface becomes
output from the isothermal regions matrix method is more controlling of the heat transfer processes shown
always more conservative than the FEA model for an for the previous directly buried case. Moving to the still
isothermal ground surface, which is the worst case air ground surface condition for the model with no
ground surface condition used in the FEA model from a water flow reduces the rating by 3.8 00.
ratings perspective. The difference in rating results
output from the FEA model and those from the Comparing the same FEA model with an isothermal
isothermal regions matrix are approximately constant boundary condition at the ground surface; with the
for all burial depths with a difference of between water cooling switched off the ratings are 42 00 lower
57 Amps and 59 Amps which represents a difference than with it flowing at 4.88 1 s-1. Figure 3 shows the
between the two rating methods for all burial depths of influence of the water flow rate on the cable rating,
2.6 o. values shown all represent water in turbulent flow; it is
seen that 4.88 1 s-' has been chosen to provide a good
With reference to Figure 2, when the water cooling is in compromise between flow rate and rating; positioning
operation the influence of burial depth is shown to be the flow rate to achieve most of the available benefit.
small, showing a change in rating of just 1 % when Parametric sensitivity analysis
water is flowing, compared with a 16 00 change with
burial depth from 384 mm to 1944 mm (dimension A of It is of interest to know which of the assigned
Figure 1) when no water cooling action is present. parameters has greatest effect on the rating. To

68
I
: 1400

a,

I>I
1800

1600

1I1200
1000

800

600
0 1 2 3 4 5

Figure 3, The influence of water flow on cable rating.

Symbol
-
Parameter
conductor OD
6
Water flowrate (I/s)

Value
58.3
7 8 9

Units
mm
1

1(
understand the relative influence of each parameter on
the FEA model a full factorial experiment was designed
using an L8 orthogonal array (Table 2) to find the
relative sensitivity of the FEA model to several
parameters. The advantage of this approach is to reduce
the number of models that need to be solved with the
additional benefit that the interactions between factor
can also be studied. The disadvantage is the possibility
of confounding one parameter or "factor" with another,
this is minimised here by assuming there is no
confounding for interactions such as 1&2 with 4&7
shown in column 3 and not placing parameters for
investigation in the same column as two factor
interactions. With reference to Table 2, which shows
the experimental design, the following four parameters
have been investigated; ground surface boundary
condition, dielectric loss, ac resistance and soil thermal
resistivity. The high and low levels corresponding to 1
and 0 in Table 2 are shown in Table 3.
Considering the "effects" row of Table 2 shows clearly
- dielectric OD 100.1 mm the impact of the factors assigned to columns 2,3, 4 and
- Outer-serving OD 117.5 mm 7 representing tan 6, the interaction of columns 1&2
thermal conductivity 400 W.m-1.K-1 and/or 4&7, ac resistance and soil thermal resistivity.
of copper
As expected the surface boundary condition has a
thermal conductivity 0.182 W.m-1.K-1 relatively small effect for this water cooled model. The
of fluid-filled paper
thermal conductivity 35.3 W.m-1.K-1 relatively large effect of column 3 representing the
of lead interaction of columns 1&2 and/or 4&7 is realistically
thermal conductivity 0.286 W.m-1.K-1 due to the 4&7 interaction not the 1&2 interaction
of polyethylene (PE) considering the small impact of ground surface
thermal conductivity 0.833 W.m-1.K-1 boundary assigned to column 1; the effect of this
of soil backfill interaction is clear in Figure 4b. To asses the relative
thermal conductivity 0.440 W.m-1.K-1 importance of the effects they are studied with respect
of coolant pipe to the change in the parameter by defining sensitivity as
- pipe ID 66.7 mm
pipe OD 89.3 mm A effect
-
sensitivity = fact (5)
A\ factor(5
C specific heat capacity 4.181 kJ.kg-l.K-1 Sensitivity results are shown in the last row of Table 3.
(water cig 20'C)
characteristic ac resistance
D pipe ID mm
dimension The largest effect is clearly related to ac resistance.
f friction factor -
This is due to the location of the loss and therefore heat
h heat transfer - W.M 2 K-1 generation in the centre of the cable, and because sheath
coefficient
k thermal conductivity - W.m -1K-1 losses are evaluated as a factor of ac losses. With this
L section length - mm result and knowledge that ac resistance will vary with
the cable conductor temperature, which has been shown
mass transfer rate - kg.s to change considerably along the length of a water
NU Nusselt number cooled cable, a further model has been solved with ac
Nusselt number for resistance as a variable of temperature. This model was
NUD the length D identical to run 6 of Table 4, but with the ac resistance
Pr Prandtl number 7.0 set as a variable of temperature. For this model the
q heat flux - W.m2 relationship for ac resistivity, p, and temperature is
Re Reynolds number 92839 simply defined as
Tb bulk temperature - K
bulk temperature
Tb, entering a section
- K p = P20 (1 + aAT) (6)
bulk temperature - K
Tb, out exiting a section where a is taken as 0.004 for copper and P20 as
TS surface temperature - K 26.6xlO-9 Q m. A more complex modeling of the
U0 phase voltage 231 kV ac resistance and losses could be sought for future
Table 1, Forced cooled circuit cable model parameters. modeling; for example using [5]. It is seen that the

69
Parameters
Run 1 2 1&2, 4 2&7, 1&7, 7 Results (A)
surface tan 6 4&7 Rac 1&4 2&4 Tr(soil)
1 0 0 1 0 1 1 0 2285
2 1 0 0 0 0 1 1 1951
3 0 1 0 0 1 0 1 1884
4 1 1 1 0 0 0 0 2230
5 0 0 1 1 0 0 1 1389
6 1 0 0 1 1 0 0 1624
7 0 1 0 1 0 1 0 1587
8 1 1 1 1 1 1 1 1341
Effects: 0.25 -51.75 49.75 -602.25 -5.75 9.25 -90.25
_aerav
1786
Normalized
sensitivity I04 l
--04-I-- 1l |
Table 2, Full factorial experiment results
0.48
04

Level 0 1 resistivity of the soil surrounding the cable group has to


Parameter impede the heat flow away from the cables; the
ground surface boundary Isothermal Still air additional heating from the higher loss condition
dielectric loss (tan 6) 0.0028 0.0035 interacts with this thermal impedance to greatly reduce
ac resistance (Rae) [Qm-1 1.277xl 0-5 2.554xl 0-5 the rating. Again in Figure 4b) an interaction between
soil thermal resistivity 1.2 2.5 tan 6 and ac resistance is seen. Again the increase in
(Tr(soil)) [m K W-1] 1_2 2_5 loss in the dielectric also impedes the heat flow away
Table 3, Parameters for investigation in Table 2 from the core due to the reduced thermal gradient within
the cable, and so an interaction is again evidenced by
model of run 6, Table 2, with a high Rqr provides a the change in gradient of the two plotted lines. Only a
rating of 1624 A, this is raised to 2283 A using a low small interaction is observable in the third interaction
Rae of 1.277x10-5. The result using the variable Rae was studied between dielectric losses, tan 6, and soil thermal
to raise the rating up to 2306 A, a further increase of resistivity, shown in Figure 4c).
1 00 over the model with a low Rae
Interaction plots CONCLUSIONS
The plots of Figure 4 show the influence that two In this paper a method to solve the heat-transfer
interacting parameters have on the rating and the problem for rating forced cooled cable circuits using an
significance of the interaction between factors. It is extension of a 2-D FEA model into the third dimension
worth note that each of the points on the plot represents is described. Parametric sensitivity analysis using this
an average of two results from Table 2. With reference model identified the significance of the ac resistance
to Figure 4a) it is seen that the ac resistance and thermal parameter. Implementation of variation in ac resistance
resistivity of the soil interact, evidenced by the with temperature showed an increase in rating of 1%.
difference of gradient for each plotted line. This can be
interpreted as showing the impact that the thermal REFERENCES
[1] D. J. Swaffield, P. L. Lewin, and M. LeBlanc,
2200 Tr: 1.2 tan d: 0.0028
-

-Tr: 2.5 <i


2200 -
- tan d: 0.0035
"Investigation into the Conservatism of High Voltage
S 2000 2000 Cable Rating Methods: A Comparison between IEC60287
C: and Finite Element Analysis," in XIVth ISH, International
1800 U 1800 Symposium on High Voltage Engineering, Beijing, China,
ID
'
1600 X+ 'U 1600 2005, pp. CD-ROM.
[2] D. J. Swaffield, P. L. Lewin, and S. J. Sutton, "The use of
1400 1400 finite element analysis modeling to improve the precision
1.5 2 2.5 3 1.5 2 2.5 3
ac resistance (ohm.m) Xlo ac resisstance (ohm.m) x 10-5 of high voltage cable ratings," in 10th INSUCON
a) b) International Electrical Insulation Conference,,
Birmingham, UK, 2006, pp. 155-16 1.
2200 | tan d 0.0028
[3] Cigre committee-21-08, "The Calculation of Continuous
| tan d: 0.0035

2000 Rating for Forced Cooled Cables," Electra, vol. 66, pp. 59-
84, October 1979.
1 1800
[4] D. J. Swaffield, P. L. Lewin, and S. J. Sutton, "A
' 1600- Comparison between 2-D Isothermal Region Matrix and
Finite Element Analysis Cable Rating Methods for Water
1400
1.5 2 2.5 3
Cooled Circuits," in 2006 ISEI, IEEE International
Tr soi (K.mNV) Symposium on Electrical Insulation, Toronto, Canada,
c) 2006, pp. 490-493.
Figure 4, Interaction plots for the full factorial [5] W. G. 21-03, "Large cross sections and composite screens
experiment of Table 2. design: Technical Brochure 272," CIGRE.

70

You might also like