Beruflich Dokumente
Kultur Dokumente
The purpose of this paper is to introduce a modified approach for the electric field computation at the surface of overhead transmission
line conductors through the Finite Element Method. The proposed strategy is based in a spatial transformation, well-known as Kelvin
Transformation, resulting in a special way to treat the unbounded domain. Unlike other similar applications of the Finite Element
Method in transmission lines, the proposed strategy aims to reduce the computational domain. This allow an accurate numerical
evaluation of the electric potential gradients, without the need of geometric simplifications of the conductors, in a reduced computational
time. Comparative results show that a more realistic treatment can provide a better understanding of how electric fields operate near
real conductors, demonstrating that the proposed technique can provide an accurate design for conductor systems in transmission lines.
Index Terms — Finite Element Method, Kelvin Transformation, Unbounded Electro-Quasistatic Field, Surface Voltage Gradient
0018-9464 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. (Inserted by IEEE.)
PB-A6 2
mapping functions. The radius 𝑎 of both circular domains must where 𝜎 𝑒 and 𝜀 𝑒 are related to the physical properties of the
be equal, so each point (𝑥, 𝑦) in the original external domain finite element medium, 𝑁𝑖𝑒 and 𝑁𝑗𝑒 are shape functions and Ω𝑒𝑖𝑛𝑡
can be represented by a unique point (𝜉, ) in the transformed
represents the eth subdomain considered in real coordinates.
domain, according to:
It is still necessary to develop the contribution terms for the
𝜉 = (𝑎2 /𝑟 2 )(𝑥 − 𝑥𝑐 ) + 𝜉𝑐 , (1) elements in the external domain. For these, the gradient of
shape functions in (5) must be modified due to the spatial
𝜂 = (𝑎2 /𝑟 2 )(𝑦 − 𝑦𝑐 ) + 𝜂𝑐 , (2)
transformation. The Jacobian matrix to this transformation can
where (𝑥𝑐 , 𝑦𝑐 ) is the central point in the internal domain (in real be developed through (1) and (2) and it is given by:
coordinates), (𝜉𝑐 , 𝜂𝑐 ) is the central point in the external 2
transformed domain, and 𝑟 is the Euclidian distance between (𝜂 − 𝜂𝑐 )2 − (𝜉 − 𝜉𝑐 ) −2(𝜉 − 𝜉𝑐 )(𝜂 − 𝜂𝑐 )
1
the central point (𝑥𝑐 , 𝑦𝑐 ) and the external point (𝑥, 𝑦): [𝐽] = [ ]. (6)
𝑎2
2
−2(𝜉 − 𝜉𝑐 )(𝜂 − 𝜂𝑐 ) (𝜉 − 𝜉𝑐 ) − (𝜂 − 𝜂𝑐 )2
𝑟 = √(𝑥 − 𝑥𝑐 )2 + (𝑦 − 𝑦𝑐 )2 . (3)
Using (6), the elementary contribution matrix terms can then
With this transformation, moving beyond the fictitious be rewritten for the external coordinate system as [10]:
boundary in the real domain, toward the infinity, means moving
𝑒
toward the center of the transformed domain. 𝑘𝑖𝑗 = − ∫Ω𝑒 (𝜎 𝑒 + 𝑗𝜔𝜀 𝑒 )∇𝜉, 𝑁𝑖𝑒 ′ ∙ ∇𝜉, 𝑁𝑗𝑒 𝑑Ω𝑒𝑒𝑥𝑡 , (7)
𝑒𝑥𝑡
To apply the Kelvin Transformation in TLs, this work
suggests imposing the virtual boundary close to the conductors, where ∇𝜉, 𝑁𝑖𝑒 and ∇𝜉, 𝑁𝑗𝑒 are the gradient of shape functions in
reducing the study domains, as shown in Fig. 1. The ground is relation to the new coordinates system (𝜉, ), and Ω𝑒𝑒𝑥𝑡
entirely assumed in the external transformed domain. This represents the domain of the finite element considered in
transformation ensures that the transformed domains satisfies transformed coordinates. In (7), the mediums are considered
the continuity conditions. isotropic and homogenous since their physical properties
remain unchanged with the geometric transformation.
Assembling the contribution terms of all mesh elements in
both domains (Ω𝑡 = Ω𝑖𝑛𝑡 ∪ Ω𝑒𝑥𝑡 ), the following linear system
of equations can be obtained:
[𝐾]{𝑉} = {0}, (8)
where [𝐾] is a global stiff matrix and {𝑉} is a vector with the
complex electric scalar potentials (𝑉1 , 𝑉2 , … , 𝑉𝑛𝑛 ) for each one
of the 𝑛𝑛 domain nodes.
Before solving (8) it is necessary to directly apply the
Dirichlet boundary condition. Besides the known three phase
electric potential phasors at conductors’ nodes (V𝐴 , V𝐵 and 𝑉𝑐 ),
the electric potential in a node located at the infinity (the center
of the transformed external domain) must be equal to zero
Fig. 1. Kelvin Transformation proposal to FEM applied in TL applications (𝑉(𝜉𝑐 , 𝜂𝑐 ) = 0∠0𝑜 ).
After the solution of (8), the electric field values can be easily
Once the outer region is transformed, both domains can be obtained through the electric potentials in the discretized
normally discretized. Although in each domain the mesh domains nodes [9].
elements will be different, the same nodal points in the circular
boundaries Γ𝑓 must be taken, since the fictitious limit is unique. III. RESULTS
The electric potentials in these coincident nodes must be equal. The proposed approach is used in the evaluation of a 525 kV
For the inner domain, the application of FEM does not Brazilian TL. It uses Aluminium Conductor Steel Reinforced
require additional adjustments. Considering the following form (ACSR) cables in quadrangular four-conductor bundles. The
of Poisson equation [9]:
PB-A6 3
(a) (b)
Fig. 4. Mesh elements: (a) Internal domain. (b) External domain.
(a) (b)
Fig. 2. 525kV Transmission Line: (a) Tower topology. (b) Conductor system’s
geometric data (disproportionate scaling).
Fig. 5. Superficial electric field at conductor 13 – Method comparison
The results of electric field at the surface of conductors are
When considering the real geometries of ACSR cables in
compared between the proposed FEM method (FEMProp), the
FEM approaches, there are appreciable errors in the maximum
traditional FEM method (FEMTrad) [7] and the classical well-
values for all TL conductors due to the behavior of the electric
established method of successive images (MSI) [2].
field in the straps. The comparative behavior in the conductor
Three different tests are performed. First, the methods are
21 (worst case) is shown in Fig. 6. Although the electric field
compared to each other considering ideal cylindrical conductors
seems to follow a same trend between methods, the limitations
in order to verify the proposed method validity. Second, true
of MSI impose considerable errors when it is obliged to
cross-section and materials are considered, in order to
consider simplified ideal conditions. It is noticed that the
demonstrate the advantage of using FEM in real cases. Third,
electric field abruptly varies along each round strand surface of
the proposed FEM is extended to the evaluation of the
a real conductor by FEM results. Another observation is the
superficial electric field in a trap wire (TW) cable, reinforcing
good consistency between FEM approaches across the entire
its pratical importance. Fig. 3 shows all the conductors
extension of the cable surface, validating the proposed
geometries tested. A homogeneous ground with a resistivity of
methodology. The results for maximum electric field in all
100.m is also arbitrarily considered in FEM approaches.
conductors are specified in Table I which shows an average
difference of 30% between FEMProp and MSI values, and an
average difference of 1.6% between FEMProp and FEMTrad
values, confirming the previous analyses.
values for a TW cable are located at the edge of its strands, runtime. This occurs because most of the finite elements are
different from ACSR cables that have highest values at the imposed near of the TL’s conductors. Consequently, it is not
center of its strands. The electric field in the middle of a TW possible to have a major reduction in the element number only
cable’s strands is also almost the same of an ideal conductor. reducing the study domains.
Another observation is that the surface voltage gradient value is
always smaller in TW cables than ACSR cables (19.86kV/cm IV. CONCLUSION
versus 21.13kV/cm, 6% lower, for the exemplified conductor). This paper presented a modified approach for the
TABLE I computation of electric fields in overhead transmission lines
MAXIMUM SUPERFICIAL ELECTRIC FIELD COMPARISON based on the Finite Element Method and the Kelvin
Transformation. This approach was evaluated in a real
Condu FEMProp. MSI Difference FEMTrad. Difference
ctor (%)
transmission line, and its results were compared with two other
(KV/CM) (KV/CM) (%) (KV/CM)
11 20.2 14.6 28.0 20.6 1.9
techniques: the classical Method of Successive Images and the
12 21.7 15.5 28.6 21.8 0.5 traditional Finite Element Method, as proposed in [6].
13 20.9 15.0 28.0 21.3 1.6 It is verified that FEM performs a more reliable modeling of
14 19.8 14.2 28.2 20.0 1.0 real cables, overcoming limitations imposed by classical
21 22.1 14.7 33.3 22.4 1.3 methods, like MSI. Although both FEM techniques showed
22 22.1 14.8 32.7 22.3 1.0 equivalent results, the computational effort of the proposed
23 23.6 15.7 33.3 24.0 1.9 approach can be smaller due to the domain reduction.
24 23.4 15.8 32.7 23.6 0.7
The increase of computational efficiency without loss of
31 21.3 15.6 27.1 21.8 2.3
32 20.4 14.7 27.9 20.6 1.3
numerical quality justify the use of the proposed technique to
33 19.5 14.3 26.6 20.4 4.7 search for optimal configurations for TL’s conductor systems,
34 21.1 14.9 29.4 21.4 1.1 favoring their corona performance.
Future work will demonstrate that this technique is also
accurate and efficient to calculate electrical parameters of TLs,
extending the interest on its practical use.
ACKNOWLEDGMENT
This work has been supported by the Brazilian agencies
CAPES, CNPq and FAPEMIG.
REFERENCES
Fig. 7. Superficial electric field at conductor 34 – Cable type comparison [1] EPRI, AC Transmission Line Reference Book: 200 KV and Above, Third
Edition: Electric Power Research Institute, 2005.
After evaluating the equivalence between FEM approaches [2] M. P. Sarma and W. Janischewskyj, “Electrostatic Field of a System of
and their importance for an accurate analysis of how the electric Parallel Cylindrical Conductors”, IEEE Trans. Power App. Syst., vol.
PAS-88, no. 7, pp. 1069-1079, 1969.
field operates in real cables, it is necessary to evaluate their [3] H. Singer, H. Steinbigler, and P. Weiss, “A Charge Simulation Method
computational performance results. Table II shows the number for the Calculation of High Voltage Fields”, IEEE Trans. Power App.
of elements and the average processing time for 30 runs of each Syst., vol. PAS-93, no. 5, pp. 1660-1668, 1974.
[4] Q. Li, S. M. Rowland, and R. Shuttleworth, “Calculating the Surface
technique in the electric field computation of the treated TL, Potential Gradient of Overhead Line Conductors”, IEEE Trans. Power
when considering each type of cable. The tests are performed in Del., vol. 30, no. 1, pp. 43-52, 2015.
an Intel i7-6500U CPU, 2.5GHz, 8GB RAM. [5] J. S. A. Sarmiento and M. C. Tavares, “Enhancement the overhead
transmission lines capacity by modifying the bundle geometry using
TABLE II heuristics algorithms”, IEEE APPEE Conference, pp. 646–650, 2016.
[6] E.M. Freeman and D.A. Lowther, "A novel mapping technique for open
COMPUTATIONAL EFFORT COMPARISON
boundary finite element solutions to Poisson’s equation", IEEE Trans
Magnetics, Vol. 24, no. 6, pp.2934-2936, 1988.
Number of Average Processing
Method Cable Type [7] K. Hameyer, R. Mertens and R. Belmans, "Computation and
Elements Time (s)
measurement of electromagnetic fields of AC high voltage transmission
Ideal 17936 (100%) 79.1 (100%) lines", AC-DC Power Transmission, IEE Conf. Publication, no 423, 1996.
FEMTrad. ACSR 60352 (100%) 446.8 (100%) [8] Q. Chen and A. Konrad, “A review of finite element open boundary
TW 49978 (100%) 234.8 (100%) techniques for static and quasistatic electromagnetic field problems,”
Ideal 14476 (80.7%) 60.7 (76.7%) IEEE Trans. Magnetics., vol. 33, pp. 663–676, 1997.
FEMProp. ACSR 53916 (89.3%) 332.0 (74.3%) [9] R. M. Matias and A. Raizer, " Calculation of Electric Field Created by
Transmission Lines by 3D-FE Method Using Complex Electric Scalar
TW 44594 (89.2%) 185.3 (78.9%)
Potential ", Appl. Comput. Electromagn. Soc. J., Vol. 12, no. 1, pp.56-60,
1997.
According to these data, there is a reduction of more than [10] Z. Tang, Q. Wang, H. Guo, et al. "A Domain Map Finite Element Method
20% in the solution time when using FEMProp compared with for Solving Open Boundary Electromagnetic Field Problem and its
Application", PIERS2009, pp. 1676-1680, Beijing, 2009.
FEMTrad. Although FEMProp causes a great reduction of the
study domains, larger improvements cannot be noticed over