CSE002

Volume N2, June 2015
Innovation
in the Power Systems
industry
Engineers and specialists worldwide exchange
information and state-of-the-art world practices
to enhance knowledge related to power systems
in CIGREs latest publication.
In this Issue, Best of CIGRE 2014 Session Papers

(Continued from CSE01)
CIGRE 21, rue dArtois, 75008 Paris - ISSN : 2426-1335
Dear readers,
with SIL (surge impedance loading), a subject which
is attracting increasing interest in recent years as designers and system planners are exploring the option
to increase the transfer capacity of transmission lines.
This is especially important for long distance interconnections, where increased SIL of the lines can reduce or even remove restrictions on power flow.
You are holding in your hands the second ever issue

of the Cigre Science and Engineering Journal (CSE).
The editorial team has done its best to put together
an attractive package of papers for you, which include
the remaining three best of the 2014 Paris session.
Interestingly enough all these papers are related with
the increasing use and penetration of renewables in
the grid. They cover permanent magnet direct drive
configurations for offshore wind farms, risks associated to the communication malfunctions in medium
voltage grids characterized by a high level penetration of Distributed Energy Resources (DER) as well
as the topic of providing controllable reactive power
to the next-higher voltage level by controlling of said
resources.
Last but definitely not least, this issue is hosting a

paper on energy storage systems optimization, this
being an attractive solution to achieve a more environmentally-friendly energy mix while maintaining
the safety of electricity supply. What is important
about this paper is, that it is the winning paper of
a student contest organised recently by the French
National Committee on the theme of Smart Grids to
promote Cigre among young academics.
The second group of papers in this issue is devoted

to high quality work from some of Cigrs working
groups, which in my view are better suited to be published in this journal than as Technical Brochures.
This is the case for the two papers on wind induced
motion of bundled conductors (excluding galloping),
namely Aeolian vibrations and sub-span oscillations,
prepared by WG B2.46, convened by Prof. Giorgio
Diana, which continue the successful publication history of Cigr on similar issues. Another paper deals
Finally, let me thank you for your interest in CSE

and encourage you to recommend this journal to
your colleagues but also consider it seriously for your
next paper.
Konstantin O. Papailiou
Chief Editor
konstantin@papailiou.ch
Cigre Science & Engineering
Contents of this issue
page
Editorial
Effects of the modularity in PMSM synchronous machine behaviour
Reliable Controllable Reactive Power for the Extra High Voltage System By High Voltage
Distributed Energy Resources
14
Security of communications in voltage control for grids connecting Distributed Energy

Resources: impact analysis and anomalous behaviours
30
Energy Storage Systems Optimization
40
Wind Induced Motion On Bundled Conductors

(Excluding ice galloping) - Part A Aeolian vibrations
46
Wind Induced Motion On Bundled Conductors

(Excluding ice galloping) - Part B Subspan Oscillations
56
Evaluation of High Surge Impedance Loading (HSIL) solutions

for increased natural capacity of 500 kV Overhead Lines
63
List of figures
Figure 1: Section of th modelled PMSG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Figure 2: Magnetic flux density comparison between no modular rotor (a) and modular rotor with 4 magnets/module (b),
5 magnets/module (c) and 8 magnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Figure 3: Comparison of magnetic flux density measured at rotor yoke between no modular rotor and modular rotor
with 4 magnets/module (a) and between no modular rotor and modular rotor with 4 magnets/module (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Figure 4: Comparison of magnetic flux density measured at gap between no modular rotor and modular rotor with
4 magnets/module (a) and between no modular rotor and modular rotor with 4 magnets/module (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Figure 5: Comparison between induced voltage obtained for non-modular and rotor modular with different number of magnets per module.. . 10
Figure 6: Cogging torque in non-modular rotor (a) and comparison with modular rotor configurations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Figure 7: Comparison of voltage (a), current (b) and power (c) obtained for the no modular rotor machine and modular
rotor with 16 magnets per module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Figure 1: Sketch of HV DER Q Provision for DSO-operated HV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 2: Sketch of HV DER Q Provision for TSO-Operated HV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 3: Research Topics Related To CQ Provision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 4: Schematic Representation of the EHV/HV System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 5: DER PQ Capability Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 6: Reactive Power Provision on Network Area Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 7: Bandwidth of Controllable Reactive Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 8: Factors Determining Reactive Power Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 9: Expected Advantage from Optimizing EHV/HV Taps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
16
17
18
18
19
19
19
20
Figure 10: Overview of Simulated Active Power Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 11: Base Case HV Wind Measured: Network Area P & Q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 12: Base Case HV Wind Strong: Network Area P and Q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 13: Network Area Q Bandwidth for Different Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 14: Loading of Transformer T2 for Different Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 15: Voltage Level: DER Optimized, EHV/HV Taps Local . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 16: Voltage Level: DER + EHV/HV Taps Optimized . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 17: HV DER PCC Grid Strength and CQ Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 18: Specific Cost of HV Controllable Reactive Power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 19: Cost Components Considered in Cost Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 20: Specific Cost of Reactive Power from EHV Capacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 21: Decision Flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
21
22
22
22
23
23
23
24
25
25
26
Figure 1: Voltage Control use case Function Input/Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2: Voltage Control use case SGAM mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 3: Voltage Control use case Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 4: Benchmark grid - telecontrol topology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 5: Benchmark grid - population distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 6: Benchmark grid regional RES distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 7: Benchmark grid RES distribution at substation level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 8: Voltage Control use case - Security Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 9: Voltage Control use case - SGIS Impact Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 10: Voltage Control use case - SGIS Likelihood Levels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 11: Voltage Control use case - SGIS Risk Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 12: Voltage Control use case - SGIS Risk Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 13: Voltage Control use case - mapping of IEC 62351 series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 14: Analysis steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
31
32
33
33
34
34
35
35
36
36
37
38
38
Figure 1: Illustration of the storage system principle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2: Financial result achieved by different storage systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 3: Flywheel simulation model. Modelling of saturation phenomena is not shown to enhance the figure clarity. Adapted from [21].. . . . . .
Figure 4: Battery model used to simulate charge and discharge cycles. Based on [22]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure 5: Evolution of according to time for flywheels (blue) and batteries (red). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
42
44
44
45
Figure A.2: Predicted maximum Aeolian vibration antinode amplitude (0-peak) as a function of frequency when the same tensile
load is applied on the subconductors and a constant low wind turbulence is considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.3: Predicted maximum strain (0-peak) on the bundle conductors at the suspension clamp
and at the spacer clamp as a function of frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.4: Maximum strains at the suspension clamp with the Diana model : when the same tensile load is applied to the subconductors
of the bundle and a constant turbulence (It < 0.07) is considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.5: Maximum strains at suspension clamp with Claren-Cosmai model : when the same tensile load is applied to the subconductors
of the bundle and a constant turbulence (It = 0.05) is considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.6: Maximum strains with Krispin model : when the same tensile load is applied to the subconductors
of the bundle and a constant, low, wind turbulence is considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.7 Cosmai hypothesis for variable wind turbulence index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.8: Maximum strains at suspension clamp with the Claren-Cosmai model : tension differential neglected and variable
turbulence (0.20 < It < 0.05) considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.9 Maximum strains at suspension clamp with the Diana model : tension differential 1D
and constant turbulence (It < 0.07) considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure A.10: Maximum strains at suspension clamp with Diana model : tension differential 10 D
and constant turbulence (It < 0.07) considered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
51
52
52
53
53
54
54
54
Figure B.1: Drag coefficient on windward rough cylinder as function of speed/Reynolds number from Wind tunnel experimental tests [8] . . . .
Figure B.2: Rough cylinders energy with respect to x/D amplitude for the three considered frequencies: experimental numerical
comparison. Due to the model scale, f = 0.5 Hz corresponds to f = 1 Hz full scale is both the static and dynamic bundle rotation angle with
respect to the wind. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure B.3: Test set-up showing towers, conductors and anemometers of the test line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Figure B.4: IREQ Measurements: Maximum peak to peak horizontal oscillations as function of mean wind speed. . . . . . . . . . . . . . . . . . . . . . . . . .
Figure B.5: Benchmark results: Maximum peak to peak horizontal oscillations as a function of mean wind speed. . . . . . . . . . . . . . . . . . . . . . . . . .
Diagram: Schematic representation of a Transmission Line in electric systems studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
57
60
60
61
63
Figure 1: Racket Tower; the drawings distances and heights are in meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Figure 2: Cross-Rope Tower (Chainette) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Figure 3: VX-Asymmetrical Tower . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Figure 4: VX-Symmetrical Tower . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Figure 5: Cat Face Tower . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Figure 6: Monopole tower . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Figure 7: Electric superficial gradients in 500kV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Figure 8: Graphic with the audible noise profiles (AN) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Figure 9: Graphic with radio interference profiles (RI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Figure 10: Graphic with the electric field profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Figure 11: Graphic with the magnetic field profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
List of tables
Table I: Input data for the analytical iterative procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Table II: Output data from the analytical iterative procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Table III: Rotor modularity options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Table IV:Torque period due to modules of the rotor.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Table I: Voltage limits observed by OPF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Table II: Simulated Scenarios and Strategie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Table III: Abbreviations used in (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Table I: Voltage Control use case - Information Assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table II: Voltage Control use case - Control Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table III: Security scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table IV: Voltage Control - security standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
33
35
37
Table I: Analytical Methods and Assumptions Used in the Study (Numerical table entries refer to supporting references) . . . . . . . . . . . . . . . . . . . .
Table II: Experimental tests characteristics and system configuration data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table III: spacer-damper data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table IV: wind power input and conductor self-damping for the analytical-analytical benchmark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Table B.1:Subspan oscillation test case data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
49
50
50
59
Table I: Minimum, Maximum and Average values of electric field on conductors surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Table II: Positive and Zero Sequence parameters of 500kV TLs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Effects of the modularity in PMSM

synchronous machine behaviour
Study committee A1
Working group A1.111
T. Arlabna, M.P. Comechbb*, M.T. Villnb, M.Garca-Graciab
ACCIONA Windpower, Spain
Centre of Research for Energy Resources and Consumption e CIRCE, Universidad de Zaragoza, Spain
a
Abstract
Permanent magnet direct drive configurations have
considerable advantages in offshore wind farms improving
reliability, longevity and lower maintenance by eliminating
gearboxes and external excitation systems. The size of the
multipole permanent magnet machine, necessary for direct
drive applications, exceeds manufacturing and transport
allowed limits. Thus, new machine concepts based on
modular design are needed.
This paper presents the study of the influence of rotor
modularity on the behaviour of a radial flux permanent
magnet synchronous generator with surface mounted
magnets and fractional slot windings. The electromagnetic
study is based on the finite element method and the solutions
has been obtained by using the software tool FLUX 2D.
The starting point of this study is a machine of 40 pole
pairs, 84 slots and an outer radius of 3.206m. The study
takes into account the number of modules and the number
of permanent magnet in each rotor module.
1. Introduction
Development of offshore wind power requires the design
of high power wind turbines to allow better use of the wind
resource at such locations.
To select the wind turbine topology, it is necessary to
consider two key factors: the ratio torque/cost and the ratio
torque/mass. The torque/cost factor is essentially economic
and the goal is to try to increase this rate to the extent possible
in order to obtain values of kW/ as low as possible. The
torque/mass ratio is critical in applications where weight
and size of the generator are design constraints. Typical
large wind turbines use horizontal axis configuration with
the generator mounted in a nacelle that sits at the top of a

large tubular tower. Therefore this ratio must be reduced as
far as possible, in order to achieve more compact and lighter
nacelle configurations.
Besides, in offshore applications the minimum maintenance
requirement it is especially important due to the difficulty
of access to the turbines for long periods of time. Gearbox
is one of the most critical components in wind turbines [1],
and so, avoiding these components improves the reliability
of the complete system. On the other hand, the increase of
wind turbine power rating becomes the direct drive option
more attractive than the geared ones [2]. For these reasons,
direct drive configuration seems particularly suitable for
offshore wind turbines.
Direct drive electric generator are characterized by low
speed and high torque and, therefore requires a high pole
number, which increases the generator dimension and
weight if conventional synchronous or induction generator
is selected. Permanent magnet synchronous generator
(PMSG) allow using small pole pitches due to the absence
of external excitation and rotor windings, and thereby they
achieve lower weight, improved thermal performance and
higher efficiency and energy yield [3]. Therefore, PMSG
direct drive configuration provides better performance,
lower maintenance and it is lighter than other options as
the induction motor with a gearbox [1] - [4].
Different design configurations of PMSG suitable for
direct drive are proposed in literature, that can be grouped
according to the directional flux paths in radial flux, axial
flux and transversal flux PM machines. According to
[4] radial flux PMSG with surface mounted magnets are
economically better choice for large-scale wind turbines
compared to the axial-flux machine.
*mcomech@fcirce.es
KEYWORDS
Direct-drive wind turbine, Finite element analysis, Modular rotor, Permanent-magnet synchronous generator
Cigre Sc
On the other hand, a high pole number with conventional

winding structures involves a high slot number. The
fractional slot winding, which has a number of slots per
pole and per phase less than one, has the advantage that it
does not require many slots although the pole number is
high, and as a result of this the iron and copper mass can be
reduced, along with the material cost and the end winding
length [5].
Slot per pole per phase

Turns per coil
32
Rotor radius
Stator slot height
4.00%
Gap
0.300%
Maximum ratio L/D

Rated Speed
Although required dimension in PMSG are smaller than

conventional generator, their size and mass, grow rapidly
with power capacity, and it is becoming a problem in terms
of capital cost, logistics and assembly [6]. For example, [7]
presents the design of a 10 MW direct drive PMSG with an
external diameter of 10 m, that exceeds the limits imposed
by transport and manufacture procedures. Hence, modular
designs become suitable for high power direct drive PMSG
configurations.
0.35
250%
10
r.p.m
Rotor material
Steel M24
Stator material
Steel M24
Conductor material
Cu
PM material
N30
Slot/tooth width ratio
1.5
Table I. Input data for the analytical iterative procedure.
This paper presents the study of the influence of rotor

modularity on electromagnetic behaviour of radial flux
surface mounted magnets PMSG with fractional slot
windings. Magnetic flux density distribution, induced
voltage, cogging torque and generated power has been
analysed. The electromagnetic study is based on the finite
element method (FEM) and the solutions has been obtained
by using the software tool FLUX 2D.
Parameter
Value
Units
Rated Power
2400
kW
Rated Voltage
2.019
kV
Rated Current
710.974
Eficiency
0.983
Frequency
6.667
Hz
PM length
29.25
mm
Section 2 of this paper present the PMSG geometry of

the non-modular PMSG taken as reference for this study.
Section 3 shows the results obtained from FEM analysis for
no load and load conditions and Section 4 summarizes the
main conclusion obtained from this analysis.
Magnet pole arc
144
Slot number
84
Stack length
1800
mm
Stator diameter
6412
mm
Rotor external diameter
6058.4
mm
2. Definition of analysed PMSG

geometry
Rotor internal diameter
5906
mm
mm
29.2
mm
Gap
Magnet height
The PMSG geometry has been developed following an

analytical iterative procedure from data shown in Table
I. This analytical procedure results in the dimensions of
the permanent magnet machine shown in Table II that
allow the modelling for finite element analysis of the nonmodular PMSG taken as reference.
Parameter
Value
Units
Rated Power (objective)
2400
kW
Rated voltage (objective)
1.32
kV
Number of phases
Pole pairs
40
Table II. Output data from the analytical iterative procedure.
3. Results of the FEM analysis

This section shows the finite element analysis carried
out in software FLUX 2D. Results obtained for the nonmodular machine described in the previous section are
taken as reference for comparison with modular rotor
machine.
Figure 1 shows a section of the modular rotor machine
modelled in FLUX 2D indicating the different parts of the

PMSG.
Figure 1 : Section of th modelled PMSG
To determine the number of rotor modules, it must be taken

into account that the designed machine has 80 poles and
each module must have the same number of magnets to
maintain machine symmetry. Hence, rotor can be divided
in different options as shown in Table III.
Magnets/Module
Modules
20
16
10
16
Figure 2. Magnetic flux density comparison between no modular rotor (a)

and modular rotor with 4 magnets/module (b), 5 magnets/module (c)
and 8 magnets/module (d).
Figure 3 compares the non-modular and the 4 magnets/

module configuration. In Figure 3a) rotor yoke magnetic
flux densities are shown. As it was expected, magnetic flux
density decreases strongly where the gap between modules
is located, but also a diminution between the second and
the third magnet of each module appears. Moreover, it
can be seen that magnetic flux density increases between
the magnets at the end of the module and the adjacent
ones and decreases in the centre of the module. In fact, it
can be observed that magnets can be grouped in pairs, as
it is observed in Figure 2b), in which the areas of higher
magnetic flux density between magnets in the ends of the
adjacent modules appears in yellow and the centre module
in red, indicating that lower values are reached in this area.
Same behaviour can be observed in Figure 2d) for 8 magnets
per module configuration showing the same behaviour in
pairs.
Table III . Rotor modularity options
3.1 Magnetic flux density distribution in no load

simulation depending on the number of modules
First we will study the influence of the number of rotor
modules has on the distribution of magnetic flux density.
Figure 2 compares the magnetic flux density obtained from
finite element simulation for each case. As it can be seen,
non-modular rotor magnetic flux density distribution is
uniform along rotor but rotor modularity affects magnetic
flux density distribution in both rotor and stator. In Figure
2 higher magnetic flux density is represented in yellow. In
modular design it must be ensured that saturation limits are
no exceeded, if so, the initial design should be modified to
prevent this from happening.
In order to facilitate comprehension of Figure 2, magnetic
flux density distributions measured at rotor yoke and at the
gap are compared in Figure 3 and Figure 4 respectively.
For modular rotor with five magnets per module, behaviour

in pairs cannot occur, being an odd number. Figure 3 shows
Figure 3. Comparison of magnetic flux density measured at rotor yoke between no modular rotor and modular rotor
with 4 magnets/module (a) and between no modular rotor and modular rotor with 4 magnets/module (b).
Figure 4. Comparison of magnetic flux density measured at gap between no modular rotor and modular rotor with
4 magnets/module (a) and between no modular rotor and modular rotor with 4 magnets/module (b).
magnetic flux density comparison between non-modular

and modular rotor design with 5 magnets per module. As
in the previous case an increase on magnetic flux density
in the magnets of the ends is observed while the magnets
in the central magnetic flux density remains unchanged
respect the non-modular configuration.
a fractional-slot tooth concentrated winding.

Figure 5 compares the induced voltage obtained for the
non-modular and rotor modular machines with different
number of magnets per module. As it was expected, the
induced voltage decreases with the number of modules. 4
magnets per module machine (20 modules) presents the
difference with 6 % respect the non-modular machine,
and 16 magnets per module machine (5 modules) 1.4
%. Regarding the selection of an even or odd number of
magnets per module, there is no appreciable difference
between the induced voltage of the machine with 4 and 5
magnets per module.
Magnetic flux density in the gap decreases in the area around

the rotor gap in both cases although there is no important
diminution in the other positions, as it is shown in Figure 4.
3.2 Induced voltage depending on the number of modules
Besides the magnetic flux density behaviour, it is especially
important to analyse how the rotor modularity affects to the
global behaviour of the machine by performing no load and
load test.
3.3 Cogging torque depending on the number of modules

One of the relevant aspects to be analysed when PM
machines are being designed is the cogging torque,
especially in low speed machines. Cogging torque is the
torque due to the interaction between PM of the rotor and
Induced voltage depends on the gap magnetic flux density

but also on the winding configuration. Designed PMSG has
Figure 5. Comparison between induced voltage obtained for non-modular and rotor modular with different number of magnets per module.
the stator slots and it is analysed by means of no load tests.

Cogging torque results in torque and speed ripple and it is
a undesirable effect.
number of slots and the number of poles of the machine. In

this case, there are 84 slots and 80 poles and cogging torque
period obtained from (1) is 0.2143. Figure 6a) shows the
cogging torque of the no modular machine for nine cogging
torque periods. In the no modular rotor machine the
cogging torque peak value is 60 Nm.
Cogging torque is position dependent and its periodicity

depends on the number of magnetic poles and the number
of stator teeth and period can be calculated as:
In modular rotor machines, gaps between rotor modules

affect cogging torque behaviour. Figure 6b) and Figure 6c)
compares the cogging torque of the non-modular rotor
machine with the obtained for modular rotor machines.
(1)
where LCM(nslots, npoles) is the least common multiple of the
Figure 6. Cogging torque in non-modular rotor (a) and comparison with modular rotor configurations.
10
In these figures, it can be observed that a new component

has been added to the cogging torque of the non-modular
machine and its amplitude and period depend on the
number of modules of the rotor. To obtain the period of
the component due to the rotor modules from, (1) must be
modified by adding the number of rotor modules:
(2)
Table IV shows the results for the different number of
modules analysed, and as it can observed in Figure 6b)
and c) these values of period agreed with those obtained
by FEM simulation. Furthermore, the amplitude of the
component due to the rotor modularity increases with the
number of modules, as it can be observed in Figure 6. These
component periods must be taken into account to avoid
torsional and fatigue stress that can damage wind turbine
structure.
Modules
LCM
(nslots, nmodules)
Period
(degrees)
20
420
0.857
16
336
1.071
10
420
0.857
420
0.857
Table IV. Torque period due to modules of the rotor.
3.4 Results of load simulation of the selected system

Figure 7 compares the results obtained on simulation of the
non-modular and modular machine with 16 magnets per
Figure 7. Comparison of voltage (a), current (b) and power (c) obtained for the no modular rotor machine and modular rotor with 16 magnets per module.
11
1997-2005 IEEE Trans. on Energy Conversion, vol. 22, No. 1,

March 2007.
pole when it is connected to a resistive load. Resistance

value has been calculated in order to produce the
generator rated power. As it can be seen there ishardly
any difference between voltage and current, 1.6 % in
both cases. Active power shown in Figure 7c) shows a
difference between active powers obtained by both nonmodular an modular machine of 75 kW, equivalent to
3.4%. Therefore, the five gaps introduced between the
modules of the rotor produce additional losses of 3.4%
in the rotor modular design.
[2] H. Polinder, F. F. A. van der Pijl, G. de Vilder, and P. J. Tavner,

Comparison of Direct-Drive and Geared Generator Concepts for
Wind Turbines, IEEE Trans. on Energy Conversion, vol. 21, No. 3,
September 2006.
[3] Li, H. and Chen, Z., Overview of different wind generator systems
and their comparisons, IET Renew. Power Gener., vol. 2, No. 2, pp.
123138, 2008.
[4] K. Ahsanullah, R. Dutta, M.F. Rahman, Review of PM generator
designs for direct-drive wind turbines, Australasian Universities
Power Engineering Conference, pp: 1 6, 2012.
4. Conclusion
[5] F. Libert, Design, Optimization and Comparison of Permanent

Magnet Motors for a Low-Speed Direct-Driven Mixer, PhD,
2004.
Current wind industry tendency is to develop increasing

rated power turbines for offshore installation. PMSG
direct drive configuration presents several advantages
for its application in offshore wind installation due to
their better reliability, longer life, improved performance,
lower weight, and lower maintenance, especially
important in offshore locations where the access to the
turbines is limited during most of the year.
[6] R. Scott Semken, M. Polikarpova, P. Roytta, J. Alexandrova,

J. Pyrhonen, J. Nerg, A. Mikkola, J. Backman, Direct-drive
permanent magnet generators for high-power wind turbines:
benefits and limiting factors, IET Renew. Power Gener, vol. 6, Iss.
1, pp. 18, 2012.
[7] A. Damiano, I. Marongiu, A. Monni, M. Porru, Design of a 10
MW multi-phase PM synchronous generator for direct-drive wind
turbines, 39th Annual conference of the IEEE Industrial electronics
society IECON 2013, pp. 5266 5270, 2013.
However, size of generator grows rapidly with power

capacity and it is becoming a problem in terms of
manufacturing, logistic and assembly. Therefore,
modular design appears as a solution for achieving even
higher power avoiding these problems.
Biographies
Teresa Arlaban Gabeiras received the M.Sc. degree in
Industrial Engineer in 2007 and she has since then been
working as an engineer in the ACCIONA Windpower
R&D Department. Currently, she is the Head of the
Research and Patents area. She has been involved in
various wind research projects dealing mainly with grid
integration issues, power curve improvements, load
mitigation strategies and drive train analysis. She is
the co-author of more than 10 papers for international
conferences and the co-author of more than 20 patent
applications related to control of wind turbines for
grid integration and load mitigation, transport and
installation systems, generation concepts and offshore
applications
This paper analyse the influence of rotor modularity

on the electromagnetic behaviour of an inner rotor,
radial flux PMSG with surface mounted magnets and
fractional slot windings. From FEM results obtained
it can be concluded that the behaviour of the modular
rotor machine is close to the non-modular rotor one
and it is closer with lower number of modules, which is
determined by the number of poles of the machine. In
the analysed machine, the additional active power losses
due to gaps between rotor modules are around 3.4 %.
A special mention should also be given to cogging torque
in modular machine, which depends on the symmetry of
the machine. Introducing gaps between rotor modules
causes no load torque components much higher (about
100 times) than the cogging torque values in the nonmodular rotor machine.
Mara Paz Comech received the M.S. and Ph.D. degrees

from the University of Zaragoza, Zaragoza, Spain, in 2003
and 2008, respectively. She has been a Research Engineer
in the electrical section of Centre of Research for Energy
Resources and Consumption (CIRCE), Zaragoza, and is
currently teaching and researching in the Department of
Electrical Engineering, University of Zaragoza, and also
collaborating with CIRCE. Her research interests include
power system modelling, FEM analysis, dynamic study,
and wind energy conversion systems.
References
[1] H. Fink, F. Devaux, B. Dolata, C. Perrier, New and innovative
Smart J. Ribrant and L. M. Bertling, Survey of Failures in Wind
Power Systems With Focus on Swedish Wind Power Plants During
12
Mara Teresa Villn received the M.S. degree in

Industrial Engineering from the University of Zaragoza,
Zaragoza, Spain, in 2003. She is a research engineer in
the electrical section of Centre of Research for Energy
Resources and Consumption (CIRCE), Zaragoza. Her
research interests include power system modelling, FEM
analysis, and wind energy conversion systems.
degrees from the University of Zaragoza, Zaragoza,

Spain, in 1989, and 1996, respectively. He is currently
a Professor of Electrical Engineering and the Area
Director of Electric Power System at Centre of Research
Energy and Resource Consumption (CIRCE), Zaragoza.
His main research interests include power system, power
system protection, electrical energy system, renewable
energy integration, lightning protections, and dielectrics.
Miguel Garca-Gracia was born in Saint-Brieuc, France,

on April 23, 1963. He received the M.Sc., and Ph.D.
13
Reliable Controllable Reactive Power for the

Extra High Voltage System By High Voltage
Distributed Energy Resources
Economic Attractiveness and Practical Implications
Study committee A1
E. Kaempf1*, M. Braun (University of Kassel & Fraunhofer IWES, Germany)
T. Stetz (Fraunhofer IWES, Germany)
H. Abele (TransnetBW GmbH, Germany)
S. Stepanescu (Netze BW GmbH, Germany)
reactive power consumption from underlying networks

and for the reactive power requirements of electrical
equipment in his own system.
Abstract
The topic of providing controllable reactive power (CQ)
to the next-higher voltage level based on controlling
distributed energy resources (DER) has recently received
increasing attention. From a theoretical perspective, the
use case of deferring or avoiding Extra High Voltage
(EHV) investment in compensating equipment by the
ancillary service provision of controllable reactive power
from High Voltage (HV) to EHV would seem within
future reach in certain cases. Despite the comparatively
higher losses related to HV DER controllable reactive
power provision, the use case is shown to be potentially
economically attractive under certain circumstances.
However, a considerable number of practical barriers
exist that are presented in a systematic way based on a
decision flowchart. Simulation results are based on a
close-to-real German EHV/HV system and measured
data. They highlight the benefit resulting from optimized
dispatch of both EHV/HV tap changer and HV DER
providing CQ.
To fulfill the task, the TSO utilizes mechanically

switched compensating devices, synchronous generators
connected to the transmission system and sometimes
Flexible AC Transmission Systems (FACTS) devices or
reactive power capability of certain types of High Voltage
Direct Current (HVDC) substations.
In certain regions of several countries e.g. UK, Ireland,
Germany, Spain, Poland a significant locational
penetration of wind and/or PV farms connected to the
high voltage (HV) system is observed already to date
or expected for the short to medium term future [1][7]. Frequently, grid codes oblige Distributed Energy
Resources (DER) to be able to provide CQ free of charge
to the HV system operator [8]-[12] . The term HV
system operator is preferred here over a choice of either
TSO or DSO (Distribution System Operator), since in a
number of countries, e.g. Switzerland, Germany, Poland,
HV systems are operated by the DSO rather than by the
TSO [13]. In many cases the CQ will be utilized for HV
system purposes, e.g. to stabilize the voltage at a weak
point of common coupling (PCC). In these cases DER CQ
will only sporadically be available for other applications.
These cases do not qualify for the investigations carried
out here. There is an increasing number of regions
however, in which the abundance of CQ is to date or is
expected to become - such that providing reliable CQ as
ancillary service from HV to Extra High Voltage (EHV)
may be considered. This is the point of departure for the
investigations carried out here.
1. Introduction
1.1. Motivation
The paper at hand focuses on reactive power exchange
at the interface of high voltage and extra high voltage
systems. Typically, the Transmission System Operator
(TSO) has among others - the task to ensure adequate
availability and dispatch of controllable reactive power
(CQ). He will compensate both for the residual of vertical
* erika.kaempf@uni-kassel.de
KEYWORDS
Ancillary Services, Compensating Equipment, Distributed Energy Resources, Reactive Power Provision
14
Beyond the deferral of investment in EHV compensators

further use cases may be realized by HV CQ provision. For
instance volt/var motivated redispatch of conventional
generation may be an issue from TSO perspective: In
this case, due to locational shortage of CQ, conventional
generators currently not in service might need to be
taken into service, and active power generation in some
other part of the system reduced. Or an existing fossilfuelled generator would need to change active power
output in order to be able to supply more CQ. In both
cases, the generators may need to be compensated by the
TSO for the change made to their planned schedule.
The largest reactive compensating devices with the

lowest losses will usually be connected to EHV. It is
therefore of particular interest to compare the economic
effectiveness of reactive power provision from HV DER
to that of EHV reactive power resources. At hand, the
EHV connected capacitor a low-cost low-loss EHV
resource of CQ - is chosen as benchmark for comparison
with provision of CQ from HV DER units.
One might expect that under these conditions HV
DER CQ cannot compete: Based on the simulation of a
close-to real German EHV/HV system it is shown that
indeed even loss-minimal dispatch of HV DER units
will frequently increase losses in the HV system. The
increase in network and DER losses expressed per
Mvarh delivered is in many cases significantly larger
than losses related to providing overexcited reactive
power from an EHV capacitor. In addition, depending
on the grid code, HV DER units may need to be paid for
parts of their reactive power provision.
The paper at hand channels the complexity of the discussed

scenario into a decision flowchart that makes transparent
how the potential area of applicability is narrowed down
by a large number of practical considerations. Differences
between the use cases are highlighted. The discussion is
accompanied and enhanced by simulation results from
a real German EHV/HV system: The influence of EHV/
HV transformer control strategy on the resulting feasible
reactive power bandwidth under constraining operating
conditions is demonstrated.
Yet, this fact is only the beginning of the analyses

carried out here: If an EHV capacitor with expected
short operating times is planned to be installed, this is
equivalent to a very high specific cost of reactive power,
i.e. a high cost per utilized Mvarh. The high cost is
due to the investment and capital costs related to the
investment in an EHV capacitor. HV DER CQ will be
the more competitive, the more it is already paid off,
i.e. reactive power can be provided at marginal cost.
When it is not requested, it does not cause any costs.
Should investment in DER e.g. retrofit be required
exclusively to produce the desired HV CQ quantity and
reliability, this competitive advantage is reduced.
1.2. Definitions and Scope

Generally, favorable and non-favorable connecting
points of DER to the power system may be distinguished:
A favorable connecting point allows to integrate the full
reactive power capability from connected resources
based on the existing EHV/HV transformer control
strategy without causing unwanted voltage or loading
conditions. Existing utility-owned fossil-fuelled or
hydro HV generators are frequently connected to such
favorable points.
The specific examples simulated focus on wind power

plants capable of providing reactive power even at zero
active power output. The results are generally applicable
to so far not utilized potentially controllable HV reactive
resources whose reactive power control capability is
financed by other use cases. Further examples are e.g.
suitable PV plants or industrial synchronous generators so
far frequently operated to maintain a fixed power factor.
The contribution at hand focuses on making available

CQ from resources previously not actively utilized for
managing the reactive power exchange, e.g. DER or
industry generators. While in the following, the term
DER is used for simplicity, the results may be transferred
to industry generators c. p.. Focus here is on provision
of controllable reactive power, i.e. on a situation where
reactive power output of HV DER is influenced online
according to the varying requirements of the EHV system
operator with the aim of producing a defined exchange
of reactive power between EHV and HV.
For the vast majority of worldwide HV systems the

discussed scenario will seem rather hypothetic: Yet,
with the observed fast expansion of installed DER - e.g.
wind capacity in many countries it may be of interest
to analyze the ICT, control infrastructure and grid code
requirements required for the discussed application.
In literature e.g. [14], [15], this topic of CQ provision

to the next-higher voltage level has so far usually been
15
The research topic of CQ provision from HV to EHV

addresses a situation in which EHV network planning /
operation is not (yet) taking into account the possibilities
available from HV resources. This may occur e.g. for the
following reasons:
treated in the context of integrating CQ from DER

in situations where the interface of two voltage levels
coincided with the interface between TSO and DSO,
Figure 1.
- HV is operated by the DSO, and therefore there is an

organizational boundary
- HV did previously not contain a significant amount
of controllable resources that were technically able to
provide CQ
- EHV has so far had such an abundance of CQ
available that there was no point in making the effort
of looking into HV resources.
Figure 1: Sketch of HV DER Q Provision for DSO-operated HV
The article at hand will thus only be of interest for

systems facing change in at least one of the above
categories: At heart, a situation is addressed, where it
shall be investigated whether enlarging the horizon of
analyses e.g. in the above-listed ways - may produce
economic benefit.
Hereafter, the effect of potential organizational
boundaries between HV and EHV system operator is
not further elaborated on, focus is on the principally
achievable potential that can be reached by coordinated
operation of the EHV/HV system.
Figure 2: Sketch of HV DER Q Provision for TSO-Operated HV
If HV is operated by the TSO situation represented in

Figure 2 - the reactive power provision process discussed
here will not be noticed as a separate process any more. If
the TSO possesses the ICT infrastructure described e.g.
in [16], reactive power provision will be integrated into
the TSO closed-loop contingency constrained optimal
volt-var control. HV network areas capable of providing
CQ could in this case be considered as individual voltage
control zones. Alternatively, they could be integrated
as sub-entities into an existing voltage control zone.
Contingency constrained volt-var dispatch the tertiary
voltage control will automatically ensure that the most
cost-effective controllable resource is activated while
maintaining sufficient reserves. For a closer discussion
of the concept of primary, secondary and tertiary voltage
control refer to [16]. Research interest in the article at
hand is in this case focusing on the question whether
it makes sense to integrate distributed controllable HV
reactive power resources into EHV tertiary voltage
control. This again is determined by how much
dependable potential can be obtained from them, and at
what cost. While costs of network losses are discussed
here, costs related to obtaining CQ from HV resources
will depend on the grid codes and possibly on additional
contracts negotiated bilaterally.
1.3. Literature Review

A comprehensive literature review on the topic may
be found in [17]. An overview of principally relevant
research questions is given in Figure 3. Kaempf et
al. [18] was the first source to provide a systematic
overview of use cases related to the topic. For any use
case analyzed it must be ensured that an economically
competitive, sufficient amount of CQ is provided at
sufficient reliability. The requirements related to the term
sufficient e.g. sufficient reliability, sufficient amount will depend on each particular use case investigated.
First research results discussing feasible amounts and
economic competitiveness are available [17]-[19]. In [20]
it is pointed out that optimal power flow based dispatch
of both DER and transformer tap changers yields relevant
reactive power bandwidths even under constrained
operating conditions, thus providing a contribution
to the discussion of required ICT infrastructure. The
question of long-term availability of HV reactive power
has not been discussed so far. Neither has the topic of
16
The paper is organized as follows: Section 2 describes

the simulated EHV/HV system. Section 3 Methodology
derives among others - the choice of simulated DER
penetration and network infrastructure. Section 4
Simulation Results presents achieved bandwidths of
CQ for different scenarios and control strategies. The
specific cost of CQ provision based on control of HV
DER is presented.
sufficient reliability been investigated systematically for

HV DER CQ provision. Both are most relevant when use
cases related to TSO network planning are investigated:
Deferring EHV capacitor investment is the comparatively
most challenging use case in terms of the requirements
for reliability, amount and long-term availability.
Section 5 starts by presenting the specific reactive power

costs related to the investment in an EHV capacitor.
Subsequently, the practical requirements related to
the use case of deferring or replaying EHV capacitor
investment by HV DER based Q provision are discussed.
2. Model Description and

Assumptions
Figure 3: Research Topics Related To CQ Provision
In the context of a comprehensive overview of all

aspects of grid integration of wind generation the topic
of providing CQ from subtransmission to transmission
level is raised in [21]. It is judged that due to the fact
that wind generators are frequently connected at weak
points of the subtransmission system, wind generators
are not able to provide any substantial contribution to
the reactive power balance at the transmission level.
The above statements were as may be seen in [21]
motivated by the investigation into the impact of wind
generators on system stability in South Australia.
The simulated system shall be systematically characterized

which allows to more easily estimate applicability of the
considerations in the readers own target environment.
2.1. EHV / HV System
A real HV system and its related EHV subset are
modelled. The EHV subset consists of a detailed part,
and the adjacent external network represented by
network equivalents. The subset of EHV modelled in
detail consists of a total of 35 substations on 380 and 220
kV levels. It is part of an interconnected transmission
system in which a minimum amount of transmission
connected synchronous generation is always online [22].
The connections to further modelled EHV nodes in
Figure 4 merely summarize the modelled complex EHV
system on a very schematic aggregation level, without
distinguishing between internal and external network.
1.4. Contribution and Outline

Based on the analysis of a German EHV/HV system the
contribution at hand takes up the concerns raised in [21]
in the following ways:
- First, it is shown that the in southern Germany
currently widely practiced local voltage control
of EHV/HV transformers will under certain
conditions indeed lead to unsatisfactory minimal
ratios of installed CQ from DER to CQ that can be
made available to EHV.
- Second it is shown that this minimal ratio may be
substantially improved by introducing optimalpower flow based computation of setpoints for both
EHV/HV transformer tap positions and DER reactive
power output.
- Third, an economic analysis compares specific costs
of EHV capacitors to the cost of reactive power
provision obtained for the simulated network.
The modelled EHV subset supplies two HV network

areas. One of these is the network area discussed here.
This HV study network area is connected to EHV via six
transformers, two of which operated in parallel.
The HV system supplies 73 medium voltage (MV)
systems. The most direct EHV access for most of the ten
HV wind farms is via EHV/HV transformers T2 and T3a,
T3b. As compared to the HV system status investigated
in [17] already, some lines considered to be shortly
before realization of an expansion in [17] are assumed
to be successfully upgraded here. The distance between
17
two EHV/HV substations, e.g. between T2 and T3 is

usually less than 55 km. Wind farms are sized between
Pn = 10 MW and Pn = 58 MW. The distance between
HV wind farms and the closest EHV coupling point is
usually less than 45 km. HV-side measurements of 15
minute averaged active and reactive power exchange
with MV systems were used. For EHV, the peak load
scenario was available and implemented. In this way
the typical planning condition of a German HV system
operator is largely replicated, who usually has one or two
characteristic network loading cases to represent EHV in
his simulations.
0.9 at nominal power output. The broad green curve in

Figure 5 indicates the assumption made here, while the
blue lines delineate the grid code requirements relevant
for the simulated system [25].
The reliable generation of reactive power assumed here
may be obtained in the following ways:
(i) By contractually agreed retrofit of existing DER
It may be worthwhile investigating integration of the
frequently existing park capacitors connected to the
park collector bus into the reactive power management
during zero active power output: This is possible, when
implemented in the context of centralized coordinated
control of EHV/HV tap changers and park capacitors, as
shown in [17]. The advantages of combining capacitors
and DER reactive power control have already been
pointed out in [26]. Beyond that, numerous possibilities
for retrofit of existing plants with FACTS devices or with
wind generators containing STATCOM functionality
exist [27].
(ii)
By deciding to change requirements for network
connection in the future, or by foreseeing such capability
when connecting new generators based on bilateral
agreements.
Figure 4: Schematic Representation of the EHV/HV System
2.2. Choice of simulated time period
(iii)
By combining different types of resources, e.g.
fossil fuelled industry generators and DER in a costminimal and reliable way.
Purpose here is to assess the potential of a partially highly

wind-penetrated HV system to provide overexcited CQ
under challenging conditions. Provision of overexcited
reactive power will usually be limited either by high
voltages or by equipment loading. Therefore, in the
investigated system, the high-feed-in low-load scenario
constitutes a particularly challenging situation. Results
are presented for a day featuring annual peak feed-back
from MV systems combined with varying degrees of HV
feed-in from wind power plants.
Especially to introduce utilization of existing DER

collector bus capacitors may significantly increase the
possibility to provide low-cost overexcited reactive
power, as discussed in [17].
2.3. CQ from HV DER

A significant variety of specifications regarding static
reactive power provision exists internationally [12],
[23], [24]. In this case study capabilities offered by
manufacturers since several years but usually not yet
installed in the field are assumed. The ability to provide
overexcited reactive power Q/Pn=0.484 from zero active
power generation to nominal power generation is
assumed, Figure 5. This is equivalent to a power factor of
Figure 5: DER PQ Capability Assumptions
18
3.2. Choice of DER penetration
It is assumed here that in reality, when utilizing DER

in the context of a volt-var Optimal Power Flow (OPF),
DER would receive voltage setpoints, see e.g. [28].
The following investigation shall assess whether the

given system is able to provide a significant amount
of CQ even under constrained operating conditions.
Ideally, the obtained results should be applicable for a
longer time span. One possible approach to solve this
problem consists of varying DER penetration, carrying
out network expansion measures if necessary, and
assessing the resulting CQ bandwidth for different DER
penetrations.
3. Methodology
3.1. Definition of Reactive Power Bandwidth
HV network areas are galvanically coupled regions that
mostly have two or more connection points with EHV.
Reactive power exchange is here studied on a network
area basis: The residual of reactive power exchange at all
network area EHV/HV connection points Figure 6 is computed once for the base case depicted in cyan
in Figure 7 - and once for the optimized case, depicted
in brown in Figure 7. The difference between the two
is the achievable bandwidth of controllable overexcited
reactive power, indicated by green arrows. In this
contribution, only overexcited CQ is analyzed, therefore
the term overexcited is usually omitted when referring
to the bandwidth in the following. The term reactive
flexibility is used as synonym to the term bandwidth
here. The consumer oriented counting system is applied
throughout this contribution.
Here, a different approach is taken. Figure 8 illustrates

the influence factors on the resulting CQ bandwidth of
any given HV system.
Figure 8: Factors Determining Reactive Power Bandwidth
CQ provision is about making use of the remaining

loading capacity, before a loading limit is hit, and making
use of the remaining voltage bandwidth, before the CQprovision-relevant voltage limit is reached. The more
frequently the system is close to some limit already in
its base case, the harder it is to transport additional
CQ. Therefore, the moment when a network expansion
measure is being planned, but has not yet been carried
out, is the most critical moment in the circle of increased
DER penetration and increased network expansion.
Therefore in this study, a DER penetration is simulated
that results in base case loading and voltage conditions
being close to permissible limits.
Figure 6: Reactive Power Provision on Network Area Level
3.3. Assess Reactive Power Bandwidth Based on

Operational Constraints: Contingency-Constrained
Optimal Power Flow
If network expansion is oriented towards hosting DER
Figure 7: Bandwidth of Controllable Reactive Power
19
MV systems. Measured values in 15 minute resolution

at the HV terminals of HV/MV transformers were
available, and one feed-in measurement from a wind
farm. The selected day features the peak annual feedback from MV.
at unity power factor, it is of particular importance to

verify the feasible reactive power bandwidth based on a
contingency constrained dispatch, Figure 9. In this way
it is ensured that
- A second scenario (Wind Strong) explores CQ

provision under severely limited grid capacity: The
measured values from MV systems are combined
with wind feed-in such that the permissible n-1
relevant loading limit of the transformer is reached
between 11:00 and 18:00. This scenario could occur
for a few hours per year, considering regionally
possible maximum coincidence of wind and PV
[29]. DER are continually being added to the system.
Network expansion measures to accommodate
increasing DER capacities tend to take several years
from planning to implementation [30]. Thus, in the
context of simulating dependable CQ provision from
HV to EHV, it is an interesting and relevant scenario
to investigate CQ provision in the event of capacity
bottlenecks.
3.4. Formulation of the optimization problem
Cost minimal delivery of reactive flexibility is simulated.
The minimum overexcited flexibility to be delivered
(QFlexox,min) is the constraint, objective being cost
minimization:
Figure 9: Expected Advantage from Optimizing EHV/HV Taps
the dispatch solution fulfills the requirement relevant for

operation of HV systems: N-1 security for loads must
be observed at each time step. This approach allows to
verify in how far the high reliability ancillary service
provision aimed at here may be practically feasible, and
at what costs.
Minimize:
Cost (t) = CostHVLineloss (t) + Cost EHVHVTrfLoss (t) (1)
subject to
If the base case investigated consists of a power system

already operated close to the permissible limits, as is
the case here, integrating additional control variables
that allow to redistribute power flows here: EHV/
HV transformer tap changers is expected to increase
the feasible reactive power bandwidth. This will be
demonstrated in section 4.
QEHVHVetwArea (t)
QFlexox,min
A loss-cost of 40 Euro/MWh was assumed, which

corresponds to the average of the average annual EPEX
spot market price Phelix Day Base of the years 2012
and 2013 [31]. Cost of transformer tap-changing is not
considered. For an analysis of its impact refer to [17].
While network planning may frequently be based

on analyzing selected loading conditions only, it is
preferred here to carry out time series simulations for the
assessment of highly reliable reactive power bandwidths.
Two scenarios are investigated:
Optimal power flow was implemented using the

heuristic optimization algorithm Mean Variance
Mapping Optimization (MVMO) [32], whose superior
performance for solving mixed-integer reactive power
problems was shown e.g. in [33]. For further details
regarding the chosen implementation of single swarm
- The first scenario is based on the measured

coincidence of HV wind feed-in and feed-back from
20
MVMO see [17]. Loading limits are assessed based on

the outcome of contingency screening: For lines, usually
either 70 % or 50 % loading limit under normal operating
conditions result. Throughout the scenarios analyzed
here, transformer T2 experiences highest loading:
The limit for T2 is determined to be 61.4 % based on
contingency analyses.
Max. permissible voltage for 110 kV
1.100 p.u.
Min. permissible voltage for 110 kV
0.982 p.u.
Max. permissible voltage for EHV side of

EHV/HV transformer
1.1030 p.u. (380 kV);

1.1305 p.u. (220 kV)
Min. permissible voltage for EHV side of

EHV/HV transformer
1.0650 p.u. (380 kV);

1.0363 p.u. (220 kV)
protection in case of an outage of T2. Such curtailment

would be a rare event in reality. Yet, to achieve highreliability CQ provision, the feasible CQ bandwidth
under this kind of condition should be verified as well, at
least in systems facing considerable DER expansion. The
residual of network area MV load (green) is assumed to
be the same for both the Wind Strong and the Wind
Measured scenario.
Table I: Voltage limits observed by OPF
Contingency analysis here focuses on the moment

immediately after a contingency, i.e. before DER voltage
control can adjust reactive power output and before EHV/
HV taps change position. For this short moment here
maximum voltages of 1.105 p.u. have been permitted,
otherwise values of Table I apply.
5. Simulation Results
Table II provides an overview of the scenarios and control
strategies whose results are discussed in the following.
Strategy
Scenario
Figure 10: Overview of Simulated Active Power Components
Figure 11 and Figure 12 highlight the resulting network

area active and reactive power exchanges with EHV. The
base case for assessing the reactive power bandwidth
consists of the reactive power exchange of the particular
scenario: the blue line in Figure 11 represents the
base case reactive power exchange for scenario Wind
Measured.
Base Case:
EHV/HV Taps
EHV/HV
EHV/HV Taps Control Variable in
Taps Voltage Voltage Con- OPF(Optimized),
Controlled & trolled (Local) together with HV
DER Q = 0
DER Q
HV Wind
Measured
HV Wind
Strong
Table II: Simulated Scenarios and Strategies
5.1. Base Scenarios: Wind Measured and Wind Strong.

Scenario Wind Measured allows to assess realistic
frequency distributions of specific loss costs of providing
CQ. Scenario Wind Strong allows to analyze the network
area reactive power bandwidth under challenging
normal operating conditions. Figure 10 features both
the measured (cyan plain) and the strong (cyan dotted)
HV wind feed-in. The strong feed-in is slightly reduced
by curtailment between 11:00 and 18:00. Curtailment is
required to make sure that lines do not get tripped by
Figure 11: Base Case HV Wind Measured: Network Area P & Q
21
normal operating conditions. This clearly highlights the

necessity to study reactive power ancillary service provision
to EHV based on network operation considerations, rather
than purely on network planning assumptions.
Figure 12: Base Case HV Wind Strong: Network Area P and Q
5.2. Analysis of Achieved CQ Bandwidths

The assumed installed DER Q bandwidth amounts to 146
Mvar overexcited. It is here aimed at providing 100 Mvar of
overexcited CQ to EHV, which corresponds to the size of
typical smaller EHV capacitors, and is equivalent to 68 % of
installed CQ. Figure 13 displays the achieved reactive power
bandwidth for the different scenarios and control strategies
implemented. The severe limitation of bandwidth for
local voltage control of EHV/HV transformer taps during
daytimes is clearly visible.
Figure 13: Network Area Q Bandwidth for Different Strategies
As expected from the analysis in Figure 9, the integration

of EHV/HV transformer tap positions as variables into the
optimization permits to achieve substantially improved
bandwidths throughout difficult operating conditions.
From Figure 14 follows that in scenario Wind Strong.
between approximately 10:30 and 18:00 the permissible
loading limit is reached. In fact, curtailment is required to
observe loading limits. Curtailment is assumed to be carried
out in steps of 10 % of installed DER capacity. This stepwise
active power reduction see Figure 10 - results in slightly
added freedom available for reactive power at certain
times compare to the red dash-dotted line in Figure 13.
If curtailment was carried out in steps of 5 %, less reactive
power could be made available, especially in the EHV/HV
Tap Local strategy.
It shall be pointed out here that the limitation in bandwidth
observed for the locally voltage controlled EHV/HV taps
in Figure 13 would not be observed if the n-1 contingency
constraint had not been imposed. By simulations it was
determined that if the post-contingency voltage limit was
chosen to be higher, e.g. 1.2 p.u., or no n-1 constraint was
imposed, bandwidths would amount to 120 Mvar and more,
while still observing the voltage bands of Table I under
Figure 14: Loading of Transformer T2 for Different Strategies
5.3. Analysis of Optimized EHV/HV Tap Control

Effects
The ability of integrated Tap + DER optimization to
22
shift power flows shall now be analyzed in more detail.

Figure 15 and Figure 16 highlight the difference in
network voltage level resulting from the two control
strategies. For locally voltage controlled EHV/HV taps
voltages remain within a band between 1.03 and 1.08 p.u.
The optimization of transformer taps allowed to make
use of the full available specified voltage band until the
lower boundary of 0.982. The upper voltage boundary
of 1.1 p.u. - Table I - is not reached, since corresponding
solutions do not observe the voltage boundaries in the
n-1 contingency calculations.
DER frequency distribution of reactive power during the

simulated 25 h is represented for each DER by a twosided bean plot in Figure 17. Each bean plot consists
of two sides (red and blue) which allows to compare
DER reactive power for the two control strategies. The
blue frequency distributions of reactive power values
are clearly displaced towards negative, i.e. overexcited
values, as compared to the red distributions that stand
for locally voltage controlled EHV/HV taps. DER are
sorted by the grid strength of their PCC, defined as the
ratio of short circuit power Sk over nominal active DER
power Pn, see [23]. Summarizing, the optimized control
of EHV/HV taps allowed to operate even those DER
at least partially in overexcited mode that have fairly
low grid strength. Bean plots have been created using
package beanplot [34] from software R.
Figure 15: Voltage Level: DER Optimized, EHV/HV Taps Local
Figure 16: Voltage Level: DER + EHV/HV Taps Optimized
Figure 17: HV DER PCC Grid Strength and CQ Contribution
23
Figure 18: Specific Cost of HV Controllable Reactive Power
5.4. Cost of DER CQ-Provision for Simulated Period
5.5. Summary
The resulting PQ-capability curve for both the Wind

Measured and the Wind Strong scenario is displayed
in Figure 18. The Base Case was chosen as reference for
reactive power provision, therefore, no cost is associated
(black dots). The specific cost of CQ provision is
computed to:
Based on optimized dispatch of both EHV/HV transformers

and HV DER reactive power output it was possible in
simulations to reliably provide at least 57 % (83 Mvar) of
assumed installed DER reactive power capability on the
day featuring 2013 annual peak feed-back towards HV
for this network area. Local voltage control of EHV/HV
transformer taps allowed to obtain bandwidths between -10
and -55 Mvar, depending on the network loading condition.
Thus, in the investigated network area, the widely practiced
local voltage control of EHV/HV taps does not qualify for
providing large quantities of CQ in a reliable way to EHV, if
n-1 contingency constraints are to be observed.
SpCQCScen i =
Term
SpCQC
CCtr,Scen i _CBase,Scen i
(2)
(QNA,Ctr,Scen i _QNA,Base,Scen i)t
Unit
Explanation
Scenario i:
i=1: HV wind as measured
i=2: HV wind strong
Ctr j
Controlled Case j:
j=1: DER Optimized., EHV/HV Tap Voltage
Controlled (Local)
j=2: DER + EHV/HV Tap Optimized
Base
Base Case:
DER Q = 0, EHV/HV Tap Voltage Controlled (Local)
Cost = Loss Cost * Losses, see (1)
QNA
Mvar
Scen i
Cost of CQ provision varies substantially depending on

network operating conditions. Values between a saving of
0.23 /Mvarh and a cost of 0.53 /Mvarh were observed.
Due to the target of achieving a bandwidth of 100 Mvar
overexcited, frequently the loss-minimization mode
of optimal power flow was not reached, and rather a
maximization of reactive power bandwidth was carried out.
The cost figures do not include the additional cost caused
by CQ delivery inside the DER parks.
/Mvarh Specific Cost of Controlled Reactive Power
To implement DER based HV CQ provision as discussed

here, it would need to be ensured that DER are not
disconnected from the system for other reasons, e.g. due
to being part of a virtual power plant that contributes to
the reduction of overfrequency by occasionally completely
disconnecting DER from the grid.
Residual of Network Area Reactive Power

Exchange with EHV, as measured at the
EHV terminals of EHV/HV transformers
Time
6. Economic attractiveness of
highly dependable controllable
reactive power from HV
Table III: Abbreviations used in (2)
Providing larger amounts of CQ under the simulated

difficult operating conditions comes at the cost of higher
losses, as may be observed by comparing the colors of
filled (EHV/HV Tap local) and empty (EHV/HV Tap
Optimized) symbols of same shape (e.g. circles).
6.1. Specific reactive power cost of EHV capacitors

At hand it is assumed that the TSO alternative to sourcing
24
Figure 19: Cost Components Considered in Cost Comparison
Figure 20: Specific Cost of Reactive Power from EHV Capacitors
overexcited reactive power from HV DER consists of

installing an EHV-connected capacitor. Principally the
TSO may have further investment alternatives at hand,
e.g. FACTS devices. These usually have the capability to
provide further ancillary services and therefore assessing
the specific cost per Mvarh provided while considering
investment and capital costs tends to be more complex.
EHV capacitors are being installed in many countries.
They are a typical source of voltage support for slow
phenomena [21]. Here the aim is to compare reactive
power cost of EHV capacitors to that caused by CQ
provision from HV DER as illustrated above.
capacitor incl. its EHV switching field, and operational

costs of 2 % of investment costs per year [35]. Loss
costs were assumed to 40 /MWh, see section 3.4.
Not considered are possible long term price increases
for operational costs and loss costs. From Figure 20
follows that if the capacitor is switched on only during
rare contingency events, the specific costs per Mvarh
delivered may become very high.
Once an EHV capacitor has been purchased, it is
characterized by extraordinarily low losses in the order of
0.35 kW/Mvar1, i.e. 0.014 Eur/Mvarh under the assumed
loss cost. Specific reactive power cost of EHV capacitors
is thus dominated by investment and capital cost.
Generally, the cost per Mvarh of reactive power delivered

is influenced by the cost components shown in Figure
19. Here, costs from the perspective of the network
operator(s) are computed, without distinguishing
whether they occur on EHV or on HV side. Generally, HV
DER CQ requires ICT & OPF integration of substantially
more components than CQ from EHV capacitors.
Therefore in Figure 19, ICT & OPF integration costs of
the EHV capacitor are not indicated separately. ICT and
OPF integration costs of HV DER CQ may constitute
a relevant cost component. They require more detailed
analysis to take into account potential synergies with
other use cases and further ancillary services and are
therefore not considered here.
With respect to HV DER CQ provision from the analyses

of the previous sections the following may be concluded:
The higher the amount of CQ available free of charge
to the DSO due to grid codes, and the higher the
requirements of grid codes with respect to providing CQ
even at zero active power output, the lower the relevant
fixed costs for making available reliable CQ at times of
zero DER active power output. Especially if capacitors
connected at the collector bus of many existing DER parks
exist and are utilized, this results in a low-loss, highly
available source of CQ. Moreover, this resource is already
paid off, i.e. beyond the potentially required upgrade of
Information and Communication Technology (ICT) and
Control Infrastructure, no additional investment and
capital costs ensue. If modern PV inverters are used, able
to provide reactive power down to 3 % of active power
output without installing additional compensating
equipment, and EHV demand for reactive power is
expected to occur only during daytimes, then utilityscale PV plant based reactive power provision may also
be a very attractive option [36].
Maintenance cost of HV DER would be in the

responsibility of the HV DER owner. However, the
higher number and different nature of components to
be managed in the context of HV DER CQ will cause
increased administration costs for the network operator.
DER CQ is assumed to be available free of charge and
potentially added losses within DER parks are not
considered.
Based on these assumptions Figure 20 shows the specific
cost of reactive power from EHV capacitors as a function
of capacitor utilization rate. The figure was compiled
assuming a cost of 3.3 Mio. Euro for the 100 Mvar
Courtesy TransnetBW GmbH: Value applies to EHV mechanically switched

capacitor with damping network of 250 Mvar operated at 420 kV. It includes
main and ancillary capacitors as well as the inductance and damping resistor.
Losses related to harmonic currents are not considered.
1
25
power in the event of a contingency of a heavily loaded

transmission line.
Summarizing, in many cases the dominating cost

component of providing CQ based on HV DER will be
the operational costs (losses, administration). Unlike
an EHV capacitor characterized by long equipment
lifetime the DER based reactive power of HV network
areas may be flexibly adjusted to the EHV requirements.
Therefore, the risk of investing into an EHV capacitor
that may not be needed any more a decade later due to
changed system operating conditions, could be avoided
by recurring to CQ from HV.
Thus, from the perspective prior to the investment into
an EHV capacitor, it seems attractive to contemplate
the HV alternative. The strong points of this alternative
in case of expected low EHV capacitor utilization
rates having been listed, the next section is devoted to
discussing the practical impediments related to such an
undertaking.
6.2. Requirements related to the use case of deferring
or replacing EHV capacitor investment
The decision flowchart of Figure 21 summarizes the
considerations: Point of departure for the investigation
is the intention to install an EHV capacitor characterized
by low expected utilization rates. The first question to
be clarified is whether the HV system has potentially
CQ resources that have not yet been considered in the
dimensioning of the EHV capacitor. If this is the case,
the conditions motivating the EHV capacitor installation
should be clearly defined, in order to derive the conditions
under which HV CQ would be utilized.
Figure 21: Decision Flowchart
A possible scenario might be to compensate for

the increase of PV-feed-in related reactive power
consumption in LV and MV systems under conditions of
high solar irradiation and low load (target condition),
see [37], [38]. This would limit relevant CQ delivery
to daylight times. Thus, modern PV plants capable of
providing reactive power support down to very low
active power feed-in would be the ideal candidate for
compensating this reactive power consumption from
MV and LV systems on HV level. This compensation
may in fact be complemented by similar action from MV
PV plants.
Step 3 is devoted to assessing the available CQ from

HV under the defined target conditions. Step 4 is of
particular importance: In many cases it may be required
to study HV CQ bandwidth under non-normal and
contingency operating conditions, in order to satisfy the
reliability requirements from EHV. The combined results
of steps 3 and 4 will determine whether the CQ that can
be provided at the required reliability level is sufficient.
This check will rule out a large number of potential
applications. During the check the CQ potential of
neighboring network areas might also be considered, if
tolerable from EHV perspective.
The most challenging outcome of this step consists of

an EHV reactive power requirement that is uncorrelated
with HV reactive power capability. Such situation may
arise when the capacitor is installed to provide reactive
Check 7 constitutes a further stepping stone for the

considered use case: If retrofit of ICT and control
infrastructure of DSO and DER is required, and possibly
26
implemented based on the heuristic optimization Mean

Variance Mapping Optimization. This allowed to analyze
HV CQ capability by taking into account the operational
constraint of guaranteeing HV n-1 security at every time
step.
also negotiation of contracts with DER owners or

introduction of a corresponding market, implementation
time may in some cases substantially exceed the time
required for ordering and installing an EHV capacitor.
Finally, the different dynamics of HV system evolvement
as compared to that of a static capacitor installation must
be considered: From TSO perspective, HV CQ would
need to be dependably available for a longer time span,
e.g. five years or more, to impact (e.g. delay) a capacitor
investment decision.
As base case todays situation in the studied area was

chosen, in which HV DER operate at unity power factor.
CQ was defined as the deviation of reactive power
exchange with EHV in the controlled case from the base
case. It was shown that it would be possible to provide a
sustained amount of 83 Mvar of overexcited CQ to EHV
even under challenging normal operating conditions
under the given assumptions. The closer the system is
operated to its limits in the base case already, the more
beneficial it is to have the added degree of freedom
inherent in centrally optimized, coordinated dispatch of
EHV/HV tap changers and DER reactive power.
Summarizing, the stepping stones sufficient quantity,

sufficient reliability, short implementation time and
sustained availability over several years may rule out the
discussed use case in many cases.
Yet, deferral of EHV capacitor investment is certainly
not the only possible use case for HV CQ: In some EHV
systems, volt/var motivated redispatch of conventional
generation may be an issue, see section 1.1. Requirements
related to this use case of avoiding volt/var related
redispatch are far lower than those discussed so far:
Rather than being available in a sustained way for years,
it is mostly sufficient if HV CQ can be forecasted with
high precision and is then reliably available in a time
span of several hours: This requirement is likely to be
met by far more systems already today, given the high
feasible forecast accuracies for active power today [39].
The required ability to provide substantial amounts of
HV DER CQ under challenging operating conditions to
EHV has been demonstrated here for a sample close to
real system.
The economic attractiveness of HV DER based CQ was

compared to that of a planned EHV capacitor installation
with low expected utilization rates: HV CQ may in
many cases come at comparatively low investment and
capital cost, since the capability to provide CQ is to a
certain degree available free of charge due to grid codes.
It has been shown that investment in HV DER CQ
can be economically competitive with EHV capacitor
investment, if it is thus possible to completely substitute
an otherwise required EHV capacitor investment.
A decision diagram has been developed highlighting the
stepping stones of utilizing HV DER CQ to replace or delay
EHV capacitor investment. The hurdles to implementing
this use case were found to be comparatively high in
many cases. One critical point is the requirement to
provide highly reliable reactive power: While change
of HV operating conditions can be largely dealt with
by integrating EHV/HV tap changers into the OPF as
demonstrated, the capability to provide CQ during times
of DER zero active power output is mostly not available
with DER installed in todays power systems.
7. Conclusion and Outlook

HV systems in certain cases possess controllable reactive
resources that are to date not fully utilized. This may
be because they have been added to the system only
fairly recently, or because in the past, EHV CQ was so
abundant that there was no need to take care of HV
resources that are comparably more cumbersome to deal
with. With increasing DER penetration and increasing
DER ancillary service capability, this trend may be
expected to increase.
Therefore those applications of the studied use case in

which DER capability to provide CQ is correlated with
the demand for CQ are most promising for short to
medium term applications: A possible application in this
context would be to compensate for the reactive power
consumption from MV or LV DER with the same primary
energy source as that of the controlled HV DER: e.g. HV
wind farms compensating for reactive power consumption
Here, the capability of a close-to-real German HV system

to supply reliable CQ to EHV has been analyzed. A
contingency constrained optimal volt-var control was
27
from MV DER, or HV PV parks compensating for reactive

power consumption from MV and LV DER. Generally, the
implementation of CQ provision from HV DER to EHV
for various use cases is easier to realize if EHV and HV are
both operated by the TSO.
The optimal combination of reactive power provision
at HV vs. reactive power provision / balancing at MV
level is to date a field of research. One advantage of CQ
provision by HV DER as compared to that by MV DER
is that their integration into the DSO ICT infrastructure
is frequently more advanced than that of MV DER, and
the HV system status is fully known to the HV system
operator, thus making OPF easier to implement.
Further use cases beyond that of delaying or replacing
investment in EHV capacitors will be studied in more
detail in future works, as well as the capability of HV
systems to provide CQ under non-normal switching
conditions or n-1 conditions, combined with voltage
control of HV DER or industry generators and dynamic
simulations. Sensitivities with respect to different
permissible short term voltage bands in the event of a
contingency could be studied as well. The feasible CQ
bandwidth in the event of reaching HV line loading
limits related to delayed network expansion may be of
interest in some networks. For systems with high DER
penetration on 230 or 275 kV level it may be of interest
to study in how far the considerations carried out here
would also apply to the provision of CQ from 230 kV or
275 kV systems to higher voltage levels.
[4]
Irish Wind Energy Association, Wind Farms in Ireland,

[Online]. Available: http://www.iwea.com/windfarmsinireland,
last accessed: 29.04.2015.
[5]
Iberdrola, Facilities Map, [Online]. Available: www.iberdrola.es,

[6]
TPA Horvath, Wind Energy in Poland 2013, [Online]. Available:

http://www.psew.pl/en/biblioteka/raporty/,
last
accessed:
29.04.2015.
[7]
TheWindPowerNet, Countries, [Online]. Available: http://www.

thewindpower.net/, last accessed: 29.04.2015.
[8]
ENTSO-E, ENTSO-E Network Code for Requirements for Grid

Connection Applicable to all Generators, version 8th March 2013,
[Online]. Available: http://networkcodes.entsoe.eu/connectioncodes/requirements-for-generators/, last accessed: 29.04.2015.
[9]
National Grid Electricity Transmission plc., THE GRID

CODE, issue 5, revision 13, [Online]. Available: http://www2.
nationalgrid.com/UK/Industry-information/Electricity-codes/
Grid-code/The-Grid-code/, last accessed: 29.04.2015.
[10] Eirgrid, Grid Code, version 5, [Online]. Available: http://www.

eirgrid.com/operations/gridcode/, last accessed: 29.04.2015.
[11] VDE FNN, Technische Bedingungen fr den Anschluss und
Betrieb von Kundenanlagen an das Hochspannungsnetz (TAB
Hochspannung), VDE-AR-N 4120, version 1.0, in force since
01.01.2015.
[12] C. Sourkounis, and P. Tourou, Grid Code Requirements for Wind
Power Integration in Europe, Conf. Papers in Energy, Volume
2013 (2013), Article ID 437674, Hindawi Publishing Corporation,
pp. 1-9.
[13] PSE, Company Information, [Online]. Available: www.pse.pl.,
Acknowledgment
[14] H. Barth, D. Hidalgo, A. Pohlemann, M. Braun, L. H. Hansen,

and H. Knudsen, Technical and Economical Assessment of
Reactive Power Provision from Distributed Generators: Case
Study Area of East Denmark, in Proc. 2013 IEEE Power Tech
Conf., June 2013, pp. 1-6.
The authors gratefully acknowledge the information

provide by DSO Netze BW GmbH and TSO TransnetBW
GmbH.
[15] N. Etherden, M.H.J. Bollen, and J. Lundkvist, Quantification

of Ancillary Services from a Virtual Power Plant in an Existing
Subtransmission Network, in Proc. 4th IEEE PES ISGT Europe,
Copenhagen, Oct. 2013, pp. 1-5.
References
[16] T. M. Papazoglu., T. Babnik, U. Bezerra, C. Cattaneo, S. Gasperic,

J. Kostevc, R. Mihalic, U. Rudez, G. Taranto, M. Testori, and J.
Vieira On Voltage and VAR Support in Electric Power System
Operation, Electra No. 272, pp. 4-17, Feb. 2014.
[1]
Global Wind Energy Council (GWEC), Global Wind Report

2014, [Online]. Available: www.gwec.net, last accessed:
29.04.2015.
[2]
UK Department of Energy and Climate Change, Renewable

Sources of Energy,
Digest of UK Energy Statistics (DUKES)
2014, [Online]. Available: https://www.gov.uk/government/
collections/digest-of-uk-energy-statistics-dukes, last accessed:
29.04.2015.
[3]
[17] E. Kaempf, M. Braun, H. Abele, S. Stepanescu, and M. Kloubert,

Reliable Distributed Energy Resource-Based Provision of
Controllable Reactive Power from High Voltage to Extra High
Voltage Systems, DER-Journal, vol. 11, n 2 (2015), to be
published.
[18] E. Kaempf, H. Abele, S. Stepanescu, and M. Braun, Beeinflussung
des Blindleistungsaustausches zwischen bertragungs- und
Verteilungsnetzen Entwicklungsoptionen, in Proc. VDE ETG
Conf., From Smart Grids to Smart Markets, Kassel, March 2015,
ref. no. 5.4.
Eirgrid Group, ANNUAL RENEWABLE REPORT 2013,

[Online]. Available: http://www.eirgrid.com/renewables/, last
accessed: 29.04.2015.
28
[19] P. Schfer, S. Krahl, H. Vennegeerts, and A. Moser, Empfehlungen

fr die zuknftige Gestaltung des Blindleistungsaustauschs an
der Schnittstelle Verteil-/bertragungsnetz, in Proc. VDE ETG
Conf., From Smart Grids to Smart Markets, Kassel, March 2015,
ref. no. 5.3.
DSO NetzeBW, 17th July 2014, [Online]. Available: http://

www.smartgrids-bw.net/uploads/media/2014_07_23_EnBW_
Verteilnetzstudie_SGP_Huber.pdf., last accessed: 29.04.2015.
[30] R. Huber, Verteilnetzausbau in Baden-Wrttemberg,
Congress Network Expansion in Baden-Wrttemberg,
Fellbach, 12th of March 2014, [online], available: https://
um.baden-wuerttemberg.de/fileadmin/redaktion/m-um/intern/
Dateien/Dokumente/2_Presse_und_Service/Veranstaltungen/
Pr%C3%A4sentationen/140312_Stromnetzausbau/Huber_
NetzeBW.pdf., last accessed: 29.04.2015.
[20] E. Kaempf, M. Braun, A. Schweer, W. Becker, R. Halbauer,

and F. Berger: Reactive Power Provision by Distribution
System Operators, in Proc. 2014 Cigr Session, Paris, ref. n
C6_302_2014.
[21] CIGRE Working Group C6.08, Grid Integration of Wind
Generation, CIGRE brochure n 450, February 2011, pp. 69 and
79-86.
[31] Bundesnetzagentur, Bundeskartellamt, Monitoringbericht 2014,

Bonn.
[22] FGH, CONSENTEC, and IAEW, Studie zur Ermittlung

der technischen Mindesterzeugung des konventionellen
Kraftwerksparks zur Gewhrleistung der Systemstabilitt in
den deutschen bertragungsnetzen bei hoher Einspeisung aus
erneuerbaren Energien, Aachen, 2012, [Online]. Available:
http://www.50hertz.com/Portals/3/Content/Dokumente/
Presse/Publikationen/Studien/Studie-Mindesterzeugung4TSO-20120120.pdf., last accessed: 29.04.2015.
[32] University of Duisburg-Essen, Mean-Variance Mapping

Optimization Algorithm, [online], last accessed Sept 2013,
available: www.uni-duisburg-essen.de/mvmo.
[33] I. Erlich, J.L. Rueda, S. Wildenhues, and F. Shewarega, Solving
the IEEE-CEC 2014 expensive optimization test problems by
using single-particle MVMO, in Proc. 2014 IEEE Congress on
Evolutionary Computation, Cancun, July 2014, pp.1084-1091.
[34] P. Kampstra, Beanplot: A boxplot alternative for visual
comparison of distributions, Journal of Statistical Software,
Code Snippets, vol. 28, pp. 1-9, 2008.
[23] G. Di Marzio, J. Eek, J. 0. Tande, and 0. B. Fosso, Implication

of Grid Code Requirements on Reactive Power Contribution and
Voltage Control Strategies for Wind Power Integration, in Proc.
2007 International Conf. on Clean Electrical Power, pp. 154-158.
[35] German Energy Agency,dena-Studie Systemdienstleistungen

2030, Berlin, Feb. 2014, [Online]. Available: http://
w w w. d e n a . d e / p ro j ek t e / e n e r g i e s y s t e m e / d e n a - s t u di e systemdienstleistungen-2030.html, last accessed: 29.04.2015.
[24] A. Ellis, R. Nelson, E. Von Engeln, R. Walling, J. MacDowell,

L. Casey, E. Seymour, W. Peter, C. Barker, B. Kirby, and J.R.
Williams, Reactive power performance requirements for wind
and solar plants, in Proc. 2012 IEEE Power and Energy Society
General Meeting, pp.1-8.
[36] E. Kaempf, H. Abele, S. Stepanescu, and M. Braun Reactive

Power Provision by Distribution System Operators, in Proc. 2014
IEEE ISGT Europe Conf., Istanbul, ref. n 0378.
[25] Association of (German) Network Operators VDN, Transmission

Code, Version 1.1, August 2007.
[37] T. Stetz, Autonomous Voltage Control Strategies in Distribution

Grids with Photovoltaic Systems, PhD dissertation, Dept. Elect.
Engineering, University of Kassel, 2013.
[26] M. Braun, Provision of Ancillary Services by Distributed

Generators, PhD dissertation, Dept. Elect. Engineering, Univ. of
Kassel, 2008.
[38] E. Kmpf, S. Schmidt, B. Walther, S. Wildenhues, R. Eggemeyer,

J. Brantl, and M. Braun, Reactive power provisioning from
MV to HV by use of reactive power capabilities of decentralized
generators, in Proc. 2013 International Congress of the Power
Engineering Society within German Association of Electrical
Engineers, ref. n 02.04.03.
[27] American Superconductor, Optimizing Reactive Compensation

for Windparks, Whitepaper, 2011, [Online]. Available: http://
www.amsc.com/, last accessed: 29.04.2015.
[28] Westnetz, Ergnzung zu Technische Anschlussbedingungen
Hochspannung, version 01.01.2013, [Online]. Available: www.
westnetz.de, last accessed: 29.04.2015.
[39] J. Dobschinski, How good is my forecast? Comparability of wind

power forecast errrors, in Proc. 13th Wind Integration Workshop,
November 2014, Berlin, 2014, ref. n 1074.
[29] R. Huber, EnBW Verteilnetzstudie, Technical Asset Management
29
Security of communications in voltage

control for grids connecting Distributed
Energy Resources: impact analysis and
anomalous behaviours
Study committee D2
Working group D2.104
G. DONDOSSOLA*, R. TERRUGGIA
Ricerca sul Sistema Energetico RSE spa Italy
Abstract
HV
ICT
IED
IP
LV
MMS
MV
MVGC
OLTC
P
Q
RES
SAS
SCADA
SGAM
SGIS
SNMP
TLS
TSO
V
VC
The paper focuses on medium voltage grids characterised

by a high level penetration of Distributed Energy Resources
and examines the risks associated to the communication
malfunctions of an ICT (Information and Communication
Technologies) architecture implementing the voltage
control function. The mostly critical segments of the
communication infrastructure are clearly represented by
the HV/MV substation networks that, in addition to local
communications towards substation control devices and
to information exchanges with the remote control centres,
have to interact with external actors, managing distributed
generators and controllable MV loads, and with MV/LV
substations aggregating controllable small size generators
and flexible LV loads. The first step of the risk analysis
shows how the impact on the unsupplied power depends
on the topology of the controlled grid, but also on the
communication networks targeted by potential attacks.
Going deeply into the threat analysis, the paper evaluates
mis-behavioural models of voltage control communications
while someone is maliciously acting through relevant
networks. The impact and threat likelihood values are
evaluated over the five scale matrix and the resulting risk
levels are mapped on the security requirements to be
satisfied by the ICT architecture, providing references
to the most relevant security solutions currently under
standardisation.
1. Introduction
The connection of distributed generation impacts the status
of the power grid such that new grid control functions need
to consider not only the classical parameters, but take inputs
also from new sources. This implies that additional ICT
networks used by new actors for exchanging information
flows with utility SCADA networks must be considered. The
resulting overall architecture interconnects a variety of ICT
entities and network segments, either owned or managed
also by third parties. In this new scenario the analysis of the
cyber security aspects of smart grid control infrastructures
deserves special attention. The paper illustrates how
the architecture aspects of both the power grid and the
information infrastructures highly influence the outcome
Nomenclature:
DER
DMS
DoS
DSO
EMS
High Voltage
Information and Communication Technology
Intelligent Electronic Device
Internet Protocol
Low Voltage
Manufacturing Message Specification
Medium Voltage
Medium Voltage Grid Controller
On Load Tap Changer
Active power
Reactive power
Renewable Energy Sources
Substation Automation System
Supervisory Control And Data Acquisition
Smart Grid Architecture Model
Smart Grid Information Security
Simple Network Management Protocol
Transport Layer Security
Transmission System Operator
Voltage
Voltage Control
Distributed Energy Resource

Distribution Management System
Denial of Service
Distribution System Operator
Energy Management System
* Giovanna.Dondossola@rse-web.it
Index Terms:
Cyber Risks, Security Requirements, Security Standards, Smart Grids, Voltage Control Communications.
30
Figure 2 introduces the actors/sub-functions of the VC use

case and shows how they can be mapped over the Function
layer of the Smart Grid Architecture Model (SGAM) plane.
The actors of the use case are placed into the Transmission,
Distribution and DER domains in the horizontal axis. The
zones in the vertical axis vary from the Market zone of the
Aggregator to the Field zone of the control functions of
the OLTC (On Load Tap Changer), Capacitor bank, DER
and Flexible Load. In the middle we have the Generation
and Load Forecast functions placed in the cell Enterprise
zone/Distribution domain. The EMS (Energy Management
System) and DMS (Distribution Management System)
control functions are in the Operation zone hosting all the
active grid operation functions. The Substation Automation
System (SAS) and the Medium Voltage Grid Controller
(MVGC) functions are located in the Station zone.
of the risk analysis steps through their application to

the voltage control in active medium voltage grids. The
Voltage Control use case is a didactic case for illustrating
the need of cyber security in smart grid applications, first
because its behaviour influences both power operation and
economy, second because of the high level inter-networking
requirements of its ICT architecture. For simplifying the
understanding of the driving principles, the qualitative
approach to the risk analysis under development by the
European Smart Grid Information Security (SGIS) working
group [1] is used as a reference method.
2. Use Case architecture

The connection of DERs to medium voltage grids can
influence the status of the whole power grid affecting
the capacity of the DSO (Distribution System Operator)
to comply with the contracted terms with the TSO
(Transmission System Operator) and directly the quality
of service of their neighbour grids. This difficulty not only
could be transferred into charges to the DSO, but it may also
impact on the TSO operation because the scheduled voltages
at grid nodes could not be observed and voltage stability
problems cannot be managed properly. In order to maintain
stable voltages in the distribution grids the Voltage Control
(VC) function is introduced whose main functionality is
to monitor the grid status from field measurements and to
compute optimized set points for DERs, flexible loads and
power equipment deployed in HV/MV substations. Figure
1 shows the input/output schema of the VC function.
Figure 2: Voltage Control use case SGAM mapping
In order to analyze the communication and security

aspects of this use case, we need to highlight the main
interactions between the actors involved. The main
control and communication components are presented in
Figure 3. The VC function is performed by the MVGC on
a node of a HV/MV substation control network. In order
to compute an optimized voltage profile the algorithm
involves communications through components inside the
DSO area, but also exchange of information with systems
outside the DSO domain.
The TSO Control Center interacts through a permanent link
between the TSO Control Network and the DSO Enterprise
Network (or alternatively the DSO Communication
Network) with the DMS in order to send the signals
triggering the execution of the voltage control optimization
cycle. The Aggregator provides the market prices and DER
operation osts to the DMS via the DSO Enterprise Network.
Also the Load and Generation forecast interact with the
DMS through the DSO Enterprise Network. The DMS
Figure 1: Voltage Control use case Function Input/Output
31
Figure 3: Voltage Control use case Architecture
communicate with the Medium Voltage Grid Controller

via the DER /Flexible loads Control Network, possibly
deploying heterogeneous communication technologies
available in different geographical areas.
forwards the information from these components to the

MVGC. In this way the control center primary substation
communications are reduced. The DMS sends /receives
information to/from the MVGC through the DSO Control
Network. The MVGC is connected through the Substation
Automation System with the Capacitor Bank and with the
OLTC in the substation network. DERs and Flexible loads
Table I and Table II report the list of the basic VC information

assets and the main steps of the control loop, respectively.
Information Exchanged
Description
Grid Topologies
Information regarding the characteristics of the grid elements (substations, loads, generators and
lines). Configuration changes of the controlled grid (grid topology reconfigurations, new DER/
load installations)
Weather Forecasts
Weather forecast, weather data
TSO Signals
Signals influencing the execution of the voltage control algorithm (e.g. changing optimization
criteria or overriding commands): Voltage setting, Reactive Power setting, Automatic Voltage
Regulator inclusion/exclusion
Generation Forecasts
Active power production plan on an hour base for a time horizon of 36 hours (36 values of active
power). Generation coefficient 0<C<1
Load Forecasts
The future load is predicted on the basis of reference loads (seasonal patterns), stochastic fluctuations, active demand effects, weather forecast, calendar day. Load coefficient 0<C<1
Energy/Ancillary costs
Costs for the modulation of active and reactive power and reward schemes
Load/DER Features
DER Nominal Power, Capability, Controllability
OLTC Measurements and States
Voltage values, Automatic Voltage Regulator included/excluded
Capacitor Bank Measurements and States
Voltage values, Reactive power values, Capacitor included/excluded
DER Measurements
Voltage values, Active and Reactive power values
Flexible Load Measurements
Voltage values, Active and Reactive power values
Grid State Estimations
Estimation of the grid current state
Capacitor Bank Set Points
Q +/-; V +/-
OLTC Set points
V +/-
DER Set points
P +/-; Q +/-
Flexible Load Set points
P +/-; Q +/Table I: Voltage Control use case - Information Assets
32
Step Name
Primary Actor
Triggering Event
Pre-Condition
Post-Condition
Generation Forecast Estimation
Generation
forecast
Periodically
New info available
New generation forecast available
Information acquisition
DMS
Periodically /
Asynchronous
TSO signal or new info
Info integrated in the data base
Forward of Forecast data
DMS
Periodically /
Asynchronous
DMS receives new data
MVGC obtains input

for the control algorithm
Grid measurement dispatch
Third party DER /

Distributors device
Periodically
Field dispatches new measurements
MVGC obtains new measurements
Forward of grid monitoring data
MVGC
Periodically
SAS has new SCADA

and DER monitoring data
DMS receives new monitoring data
Execution of control
voltage algorithm
MVGC
Values out of range
The state is not acceptable
Computation of new setpoints
Set Setpoints
SAS / MVGC
New setpoint
New setpoints computed
Devices change their settings
Table II: Voltage Control use case - Control Steps
parameters: the geographical extension of the area of interest

(multi-nation, nation, region, province, city), the countryspecific regulations involved, the grid and DER size, the
level of RES (Renewable Energy Sources) penetration,
the population density, the grid topology (i.e. the number
of HV/MV substations, MV loads, MV/LV substations,
generators, storage devices) and the telecontrol network
topology (i.e. the number of control centres, the number of
substation links per centre and the number of DER links
per substations).
The full template reporting a step by step analysis of the VC

use case control loop is available in [2].
3. Benchmark grid and security

space
In addition to the architecture details of the use case the
real grid scenario under risk analysis has to be set. The
definition of a grid scenario includes the following grid
Figure 4: Benchmark grid telecontrol topology
Figure 5: Benchmark grid population distribution
33
amount of renewable power varies according to the

estimated distribution in Figure 6. Assuming an uniform
RES penetration in all the substation of a given centre, the
estimated RES power per substation in the Italian centres
may be distributed as shown in Figure 7.
As for the RES penetration a realistic 2020 scenario

installing 40GW of renewables in the Italian medium
voltage grids [3] is used in the analysis. According to the
Italian territorial configuration, the geographical area of
the benchmark grid covers 19 regions served by thousands
of primary substations controlled by 29 centres. By taking
into account the 2020 planned grid capacity, the estimated
number of substations per centre is shown in Figure 4, while
the relative distribution of registered population per centre
is shown in Figure 5.
The effect of attacks to the telecontrol network of the

benchmark grid depends upon the security (i.e. integrity,
availability, confidentiality and non-repudiation) scenarios
in the scope of the analysis, where a given security scenario
is characterized by the parameters in Table III.
According to the Italian grid code and the national

connection rules, the size of renewable generators that
have to be mandatory connected to the medium voltages
falls within the power range of [0.2, 6]MW. Depending
upon climate conditions in the Italian regions, the targeted
By instantiating the security space on the specific VC

network topology and information assets, our use case
security space (Figure 8) covers the security scenarios
reported in the third column of Table III.
Figure 6: Benchmark grid
regional RES distribution
Figure 7: Benchmark grid

RES distribution at substation level
34
Parameter
Description
Voltage Control Scenario
Attack Target
Network interface targeted by the attack
DER interfaces,
substation2DER interfaces,
substation2center interfaces,
center2substation interfaces
Attack effect
Loss of messages (availability);

insertion of fake messages (integrity)
loss of inputs to the VC algorithm, loss of output set points

fake inputs to the VC algorithm, fake output set points, faked monitoring data
Attack extension
# network interfaces under attack
# DER networks
# substation networks
# center networks
Data frequency
Periodic / Asynchronous
periodic and asynchronous VC inputs/outputs

Table III: Security scenarios
security space. A risk level for each information asset/

security scenario can be obtained crossing the related
impact and likelihood levels.
The impact of attacks is evaluated through the five-scale
impact matrix in Figure 9 defining the levels of operational,
financial and additional risks. From the application of the
SGIS impact levels to the benchmark grid, the operational
Risk Impact Levels can be assigned to the information assets
of the VC use case. Lets evaluate the operational risks starting
from the Energy Supply risk category (leftmost column in
Figure 9). By focusing on the extreme case analysis, i.e. on
those grids in those regions with maximum DER penetration
(see Figure 6) and highest power demand (Figure 4) and on
integrity scenarios introducing fake messages causing loss
of loads, generators disconnections or substation trips, the
loss of energy supply varies with the attack target and the
related information assets: in the case of substation2DER
interface attacks the loss may be up to 100MW (yellow circle
in the picture), in the substation2centre interface attacks it
may be up to 1 GW (orange circle), in the centre2substation
interface attacks it may be up to 6GW (red circle). As for the
Figure 8: Voltage Control use case - Security Space
4. Risk analysis
According to the SGIS risk analysis process [1] the evaluation
of the risk levels of a given smart grid use case goes through
the application of the impact and threat likelihood analysis
to the scenarios of the use case information assets in the
Figure 9: Voltage Control use case - SGIS Impact Levels
35
Figure 10: Voltage Control use case - SGIS Likelihood Levels
authorization decision. For this reason an authentication

mechanism is associated to each role. The access control
matrix assigns to each data type for each (user or service)
role specific rights (read, write, update, delete). In the VC
use case, for example, only MVGC has the right to write set
point to DER. Furthermore it is important aspect to take
into account is the number of users/services for each role.
impact of such attack effects on the registered population,

the voltage control use case falls into the Medium level,
while the impact on critical infrastructures may be high or
critical, depending on the presence of essential or national
infrastructures in the sub-regions.
Now the likelihood of threat/attack occurrences actually
causing those effects has to be estimated. The level
of likelihood is evaluated for every information asset
considering parameters such as threat sources/actors, their
motivations and capabilities to achieve an attack effect
through compromise methods and in presence of essential
security counter-measures. Examples of threat sources/
actors are: disgruntled employees; vandals; amateur or
professional hackers; commercial competitors due to
industrial espionage; political pressure groups/activists;
organised criminal groups and terrorists.
Another step of the threat analysis considers the

architecture characteristics as type of services running on
the components, technologies and implementation aspects.
In the VC use case the basic components are the different
IEDs (Intelligent Electronic Devices), servers and routers
(at different DSO subdomains as ICT maintenance centre,
control centre or substation). For each of them it is necessary
to consider the configuration parameters of software layers/
modules, for example the operating systems and protocols
used for implementing the communications. In the VC
use case we suppose that the servers run a UNIX based
Operating System. DER-substation and intra-substation
communication uses IEC 61850 over MMS and for
substation-center information flows the IEC 60870-5-104
standard protocol is used. They both are connection-based
flows supported by the TCP/IP reliability mechanisms.
The VC use case comprises heterogeneous network
technologies. The center-substation links usually deploy IP
based wired networks, whereas the substation-DER links
can use wired as well wireless networks.
Both the identification of threat actors and the evaluation

of their threat capabilities to compromise the information
assets may be driven by the analysis and management of
roles in the control application. By focussing on the DSO
domain of the VC use case, we may have several user/
service roles for grid operation and ICT maintenance that
could become threat actors. Examples of possible user roles
are: local operator, remote operator, normal ICT user, ICT
administrator, ICT Security administrator. Examples of
possible service roles are: DER controller, MVGC, SAS and
DMS. Each role defines a trust level and it is used to take
Figure 11: Voltage Control use case - SGIS Risk Calculus
36
Figure 12: Voltage Control use case - SGIS Risk Levels
Behind these structural aspects the knowledge about

the control loop behaviour, as reported in Table II, is
essential for building effective attack processes whose
actual effectiveness also depends on the data frequency.
For example the success of DoS attacks, such as flooding,
buffer overflow and, resource exhaustion, will be higher
on periodic information flows (i.e. measurements and
monitoring data) then on asynchronous information flows
(i.e. setpoints).
communications (from a medium impact level to an high

risk).
By grouping the VC use case information assets and attack

scenarios considering similarity in their parameters, we
identify three main categories of assets according to the
attack target interfaces and five most relevant attacker
profiles. By applying the SGIS five scale likelihood levels in
[1], the analysis described above identifies for the VC use
case the threat levels presented in Figure 10.
5. Security standards
Figure 11 represents a numerical approach for the

calculation of risk levels pro-posed by SGIS, where the
qualitative values of impact and likelihood are summed.
These numerical values are mapped through the matrix in

Figure 12 where the risk (security) levels are identified.
Qualitative approaches as the SGIS toolkit provide only

a rough estimation of the risk value for each assets and
scenario. In order to obtain more precise evaluations the
application of quantitative risk assessment methods is
envisaged.
A set of security requirements have to be associated to the

use case risk levels. With reference to the NIST requirement
classification in [4], the following groups of security
requirements can be identified as relevant to the VC use case
assets/scenarios achieving the high and critical risk levels:
Combing the VC impact levels (Figure 9) with the

likelihood levels (Figure 10) by means of the SGIS risk
matrix, the High and Critical risk levels are identified for the
VC use case, depending on the information assets/security
scenarios under consideration. To be noticed that the
combination of the impact with the likeli-hood analysis has
increased the need of security protection of substation-DER
Access Control (SG.AC)

Identification and Authentication (SG.IA)
Smart Grid Information System and Communication
Protection (SG.SC)
Smart Grid Information System and Information
Integrity (SG.SI)
Cryptography and Key Management.
In order to meet the Voltage Control use case security

requirements, the list of security measures from technical
standards in Table IV can be selected. To be noticed that
the maturity level of the selected standards varies from
available international standard to technical specifications
or committee drafts.
Standard Type
Standard Reference
Communication protocol security standards
IEC 62351 Parts 3/4/5/6
Network security standards
IEC 62351 Part10
Role-based access control
IEC 62351Part 8
Key and certification management
IEC 62351Part 9
XML security
IEC 62351Part 11
Enabling standard IT security protocols
TLS
IPSEC
SNMP
https
ssh
Table IV: Voltage Control - security standards
37
Figure 13: Voltage Control use case - mapping of IEC 62351 series
security analysis process considering a couple of the assets

identified during the analysis.
Figure 13 depicts where the different parts of IEC 62351 have

to be applied according to the communication protocols of
the VC use case. Depending on the risk levels of the related
information assets, more or less costly implementations of
the security measures, suck key management grid/network
monitoring, will be deployed.
6. Conclusions
From the analysis of the security aspects of the voltage
control communications in active medium voltage grids the
paper described how the ICT architecture, the benchmark
grid data and the security space are involved in the impact
and threat likelihood evaluations. Through the application
of the SGIS risk analysis method, the paper showed the
advantages and the limits of a qualitative approach to the
complex task of evaluating and treating cyber risks of the
smart grid applications. Finally the paper clarified the
Starting from the information assets related to the use case,

the impact and likelihood levels are considered in order to
obtain the risk levels. The output of the risk analysis is used
in order to identify the high level NISTIR requirements
connected with the VC function security aspects. Taken
into account the set of requirements and the output of the
risk analysis, the more appropriate technical standards are
selected. Figure 14 summarizes two examples of the overall
Figure 14: Analysis steps
38
link between the outcome of the use case risk analysis and
the security requirements/measures from available and
ongoing security standards.
[4] Smart Grid Interoperability Panel Introduction to NISTIR 7628

- Guidelines for Smart Grid Cyber Security, Technical Report of
the Cyber Security Working Group, Sept. 2010. [Online]. Available:
http://www.nist.gov/smartgrid/upload/nistir-7628_total.pdf
7. Acknowledgment
9. Biographies
The research leading to these results has received funding

from both the EU 7th Framework Program (FP7/20072013)
under grant agreement no 318023 for the SmartC2Net
project, and by the Ministry of Economic Development
with the Research Fund for the Italian Electrical System for
the project Active Networks and Distributed Generation.
Giovanna Dondossola is Senior Scientist at the T&D

Technologies Department of the RSE research company
where she technically manages and leads national and
European projects on the evaluation of cyber risks in
energy grid communications. She founded the RSE
ResTest Lab for resilience testing of grid control networks
deploying redundant technologies and security solutions.
She is an active member of several international working
groups on power system information security (Cigr, CEN/
CENELEC/ETSI), and of the IEC TC57 WG15 developing
security standards for power system communications.
She is author of about 100 publications and lectures in
international conferences, journals, books and courses.
8. References
[1] CEN-CENELEC-ETSI Smart Grid Coordination Group Smart
Grid Information Security, Technical Report of the Smart Grid
Information Security Working Group, Dec. 2012. [Online].
Available: ftp://ftp.cen.eu/EN/EuropeanStandardization/HotTopics /
SmartGrids/Security.pdf
Roberta Terruggia obtained her Ph.D. degree in Computer

Science from Universit degli Studi di Torino, in 2010
and she is currently Researcher at the T&D Technologies
Department of the RSE research company where she is
involved in national and European projects on the cyber
security of energy grid communications. Her current
research topics cover studies on cyber-power risk analysis
and experimental activities in the RSE ResTest Lab for
protecting grid control networks from cyber attacks. She is
author of several publications and lectures in international
conferences, journals, books and courses.
[2] G. Dondossola, R. Terruggia, S. Bessler, J. Grnbk, P. Zwickl, R.

Lvenstein Olsen, F. Iov, Ch. Haegerling, F. Kurtz, D. Iacono, A.
Bovenzi, S. Marzorati, A. Carrapatoso, SmartC2Net Use Cases,
Preliminary Architecture and Business Drivers, SmartC2Net
European Project, Deliverable D1.1 version 2, Sept. 2014.
[Online]. Available: http://www.smartc2net.eu/public-deliverables/
smartc2net-use-cases-preliminary-architecture-and-businessdrivers/@@download/document/SmartC2Net_D1.1_v2.pdf
[3] P. Petroni, Smart Grids Operation, automation and protection
issues presented at the CIRED Workshop on Integration of
Renewables into the Distribution Grid, Lisbon, Portugal, 2012.
39
Energy Storage Systems Optimization

CIGRE national committee France
Guillaume Mancini1, (INSA Lyon France),

Pierre-Alexandre Levy1, (INSA Lyon France)
supply. As regards demand the commission advocates

the development of variable electricity prices in order to
encourage customers to shift their demands to off-peak
hours. In France, actions are taken in order to anticipate
European legislation by using smart meters. By allowing
the collection of detailed information on the demand, by
managing the reduction of consumption during peakhours, and by setting variable electricity rates, these
meters should introduce important improvements in the
management and optimisation of the electric grid [5].
Abstract
Energy storage is an attractive solution to achieve a more
environmentally-friendly energy mix while maintaining
the safety of electricity supply. In this article, a control
algorithm for energy storage systems management is
proposed. It relies on a control law that interfaces a power
plant, the electric grid and a storage system. Designed
for producers, this law enables the generation of a
profit by selling a greater part of the production when
market prices are high. From a grid perspective, this
allows a better match between production and demand
by providing a greater amount of power during peakhours. The law was designed to be applicable to a broad
range of scenarios, while being simple enough to quickly
optimize the equipment and its size. In order to illustrate
the control law operation, a case composed of a wind
power plant and an energy storage system constituted of
batteries and flywheels is considered.
The challenge of high level of REs penetration in the

electric grid remains. Indeed, France will most likely aim
at a final gross consumption of RE of 23 % by 2020, and 32
% by 2030 [6]. It seems plausible that the new organisation
of the electricity market, structured by the Nome law
in France [7], will increase the variability of electricity
prices. Electricity prices should reflect more accurately
the production costs that are variable during the day and
fluctuate as a function of the marginal production price.
Throughout this paper, energy storage is proposed as a
potential solution for grid stabilisation, as well as a means
to optimise the electricity producers profits. These two
notions are linked: on the one hand the producer will be
able to sell at the most profitable time, and on the other
hand the grid will benefit from extra-power during peaktime
1. Introduction
Driven by the European Union, the liberalization of the
electricity market has begun based on the directives 96/92
[1] and 2003/54 [2]. Since July 1st 2007 the electricity
market in France has been opened to competition by
integrating free-price offers [3]. More recently, the
European commission has released a communication
report presenting the major future orientations for the
grid [4]. The progressive cessation of an advantageous
rate for Renewable Energy (RE) purchase is advised. The
objective is to suppress national support as technology
becomes mature enough, i.e. a gradual exposition of RE
to the market price. The final objective of this incentive
is a reduction of the cost for consumers. The problem of
RE integration, which is an issue of grid stability, must
be solved by a better match between demand and supply.
As regards production, the commission proposes the
installation of back-up thermal units, shared at European
level, in order to compensate for the irregularity of RE
An algorithm has been implemented to both manage

Energy Storage Systems (ESS) and optimise producers
profits. At each instant, this model chooses the best
option between the immediate resale of the production
to the grid, and the storage of the production in one or
several Energy Storage Technologies (ESTs) in order to
defer the resale. The solution proposed is sufficiently
general to account for various scenarios while being
simple enough to allow a quick optimization of the size of
the installation. In order to illustrate its operation, a wind
power plant associated with a storage system constituted
of electro-chemical batteries and flywheels is modelled
below.
*mancini-guillaume@laposte.net
KEYWORDS
Control law, Energy storage system, Optimization, Renewable Energy
40
Fig. 1: Illustration of the storage system principle.
2. System Description
ESTs can be found in [11], [12]. In the present scope,

it has been chosen to focus on technologies for which
industrial maturity has been demonstrated. Furthermore,
Pumped Storage Power Station (PSPS) and Compressed
Air Energy Storage (CAES) induce significant disruptions
for their constructions and can only be deployed under
heavy technical and geographical constraints. Even
though these technologies have proven to be mature, this
article is focused on a case where the producer does not
have any access to these technologies. Hence, it has been
chosen to work on a storage system constituted of electrochemical batteries and flywheels.
2.1 Global description

The investigated system is constituted of three main
elements (see Fig. 1)
A wind power plant that supplies variable power
depending on wind conditions.
Some ESTs with different characteristics.
A calculator that manages the plant power flows
and acts as a power distribution system. Inputs of
this system are the power coming from the wind
turbines, and the states of the different ESTs. In
output, the repartition system supplies power to the
EST or to the electrical grid.
Lead-acid Batteries
The Valve Regulated Lead Acid (VLRA) technology
with a gel electrolyte has been retained for the lead-acid
batteries. Indeed, this technology shows a longer lifetime and a greater capacity than the Absorbent Glass
Mat (AGM) [13]. Even though this battery is not the best
battery available in terms of performance, it is mature
and relatively cheap [14]. The simulation was performed
using deep cycle gel batteries from Victron energy (ref
BAT412201100).
2.2 Production data

Wind production data have been provided by EDF
subsidiary: Rseau de Transport de lElectricit (RTE)
[8]. These data represent intra-day forecasts, i.e. forecasts
realised for the coming hour and thus constituting a good
estimate of the real production. It is important to note
that these data incorporate the data from all French wind
power plants and hence, they represent a national average
of the wind production.
Flywheels
Flywheels are constituted of a spinning disc that stores
the energy mechanically. Recently, flywheel technology
has greatly improved and the most recent flywheels are
constituted of a composite disc that can rotate up to
100,000 RPM. In order to reduce the drag losses, the disc
levitates above a permanent magnet in a low-pressure
environment. An electric motor-generator is used to
convert electric energy to mechanical energy by applying
a torque on the disc.
2.3 Sales price data

Todays legislation evolution tends to indicate that
producers are likely to sell their electricity at a price
converging toward the wholesale electricity market in
the future. The most reliable prices available for the spot
electricity market price are provided by EPEX SPOT
(European Power Exchange) [9]. It was chosen to work
with the Day-Ahead Auction prices of EPEX in order
to allow the calculator to anticipate the evolution of the
electricity exchange prices one day in advance [10].
The main advantages of these systems are their small

response-time, their extensive life-time and excellent
efficiency in charge and discharge (round-trip efficiency
ranges from 85 % to 95% [15]-[16]-[17]). The main
limitations of this EST are their relatively high costs and
the remaining friction losses that decrease the energy at
the rate of 1 to 3 % per hour. It is a mature technology
that has already been used in industrial applications [18].
2.4 Storage technologies selection

Multiple storage technologies exist: they differ in cost,
maturity, capacity, efficiency, etc. Literature reviews that
provide good overviews and comparisons of the different
41
3. Control Law Formulation
The control law relies on the concept of gain. The gain

is defined as the difference between the resale of the
production in the future as opposed as the immediate resale.
The deferred resale of the production is made possible by
the use of an EST , characterized by its efficiency. The gain ,
computed at a time , is then expressed by (3):
The control law defined below aims at managing the

power repartition between a storage system and the
electric grid. Several similar laws have been proposed in
the literature [19], [20]. The main strengths of the present
law are its generality, so that various strategies of storage
can be investigated, without sacrificing its simplicity in
order to allow quick evaluation of these strategies.
3.1 Characterization of the storage technologies
Where E(to) is the energy stored during a duration and

C(t) is the resale price of electricity at time t.
In order to characterize the EST investigated, models

from the literature have been chosen and modified for
the need of the simulation:
3.3 Control Algorithm

The simulation is divided in time-steps of duration .
At each time step, the algorithm browses the different
couples (EST k; storage duration ) that are feasible and
computes the gains G associated. Then, the repartition
system fills the different EST by favoring the most
profitable couples. When no profitable couple remains,
the rest of production is sold to the grid.
For flywheels, the model has been retrieved from

the work of Azubalis et al. [21] (Appendix A Fig. 3).
The model of Tremblay and Dessaint [22] has been
chosen for lead-acid batteries (Appendix A Fig. 4).
These models simulate storage operation with high
fidelity but they require a significant computational time.
In order to obtain a simplified behaviour that could be
used by the control law, general parameters have been
retrieved from these models. These parameters are the
maximum energy that can be stored Emax, the maximum
power that can be absorbed by the storage Pmax and a
parameter representing the efficiency of the technology:
. This last parameter indicates the proportion of energy
that can be retrieved and delivered to the grid after a
duration of storage, as defined in (1):
In order to improve the financial result, it has been

decided:
To prefer short-term storage operations in order to
increase the number of cycles per day
To favor the charge of the EST during the periods
when electricity is the cheapest
3.4 Results
A production time of Tsim=8 days, divided in time-steps
of duration =10 min, has been modelled. The financial
result is plotted in Fig. 2.
E(to) is the energy injected in the storage at time to,

while E(to+) is the energy that can be retrieved from the
storage at time to+. The parameter is calculated for
the two ESTs investigated by simulation of the models
exposed above (Appendix A, Fig 3, 4). After conducting
the analysis (Appendix A, Fig 5), flywheels seem to be
more interesting for short cycles since they benefit from
excellent round-trip efficiency. Lead-acid batteries can
be used advantageously for long cycle (i.e. several hours)
due to their very low self-discharge rate.
3.2 Definition of the variables computed
At the end of the simulation, the income I of the electric
production is computed. It represents the quantity
of energy produced weighted by the retail price. The
financial result R of the power plant associated with the
storage system solution is then calculated using (2):
Fig. 2: Financial result achieved by different storage systems.
The financial result increases proportionally to the storage

capacity as long as this capacity remains limited. After a
certain threshold, increasing the storage capacity does
42
not yield any further gains: a saturation phenomenon

appears. It is interesting to notice that this saturation
phenomenon begins for a storage capacity close to 24
hours of production, which matches the cyclic nature of
the retail prices. The financial result is maximized by using
a mixed storage system rather than a single EST. Indeed,
having access to different technologies allows one to take
advantage of their respective assets. On the investigated
case, the following conclusions can be drawn:
Further, a subsidy system could be imagined in order to

remunerate the service brought to the grid by making the
production match the demand. These kinds of incentives
have already been experienced, for instance for companies
reducing their consumption during peak-hours.
From this perspective, we see a real opportunity for
alternative storage technologies in the medium term. An
intelligent management of the storage system will be one
of the keys for the development of such solutions.
The financial result is proportional to the storage

capacity. The different storage systems investigated
exhibit very similar results: the selection must then
be made with respect to economic criteria.
5. References
After a certain threshold, a saturation phenomenon

appears: availability of different ESTs is a way to
further increase the financial result.
[1]
EuropeanParliament,Directive96/92/ecoftheEuropeanParliament
and of the Council concerning common rules for the internal
market in electricity, Available online at: http://eur-lex.europa.eu/
legal-content/en/TXT/HTML/?uri=CELEX:31996L0092, 1996.
It should be stressed the results are applicable only to

the case we investigated. Nonetheless, the control law
proposed in this paper is sufficiently general to study a
wide range of scenarios, for instance different power
production profiles, different ESTs, shorter or longer
periods of simulation, etc. The control law could then be
used to evaluate different storage strategies and is a useful
tool to calculate a return on income.
[2]
European Parliament, Directive 2003/54/ec of the European

Parliament and of the Council concerning common rules for the
internal market in electricity and repealing directive 96/92/ec,
Available online at: http://eur-lex.europa.eu/legal-content/en/
TXT/?uri=CELEX:32003L0054, 2003.
[3]
Ministre de lcologie du dveloppement durable et de lnergie, Les

tapes de louverture la concurrence du march de llectricit,
Available online at : http://www.developpement-durable.gouv.fr/
Les-etapes-de-l-ouverture-a-la.html, 2011.
4. Conclusion
[4]
European commission, Communication from the commission:

Delivering the internal electricity market and making the most of
public intervention, Available online at: http://ec.europa.eu/energy/
sites/ener/files/documents/com_2013_public_intervention_en_0.
pdf, 2013.
[5]
Ministre de lcologie du dveloppement durable et de lnergie,

Les amliorations permises par Linky, Available online at : http://
www.developpement-durable.gouv.fr/Les-ameliorations-permisespar.html, 2014.
[6]
Assemble Nationale, Projet de loi relatif la transition nergtique

pour la croissance verte, Available online at: http http://www.
assemblee-nationale.fr/14/pdf/ta/ta0412.pdf , 2014.
[7]
Assemble Nationale / Snat, Loi portant sur la Nouvelle

Organisation du March de lElectricit (NOME), Available online
at: http://www.legifrance.gouv.fr/affichTexte.do?cidTexte=JORFTE
XT000023174854&categorieLien=id, 2010.
[8]
Rseau de Transport de llectricit, Actual generation per

production type, Available online at: https://clients.rte-france.com/
lang/an/visiteurs/vie/prod/realisation_production.jsp, 2014.
[9]
European Power Exchange, EPEX spot in the power market,

Available online at: http://www.epexspot.com/en/company-info/
epex_spot_in_the_power_market, 2014.
The main contribution of this article is the formulation

and implementation of a control law that allows a rapid
evaluation of a storage system performance. This control
law is sufficiently general to be used with all types of
storage (PSPS, electrochemical batteries, flywheels,
CAES, etc.) and this with only a few parameters to
define.
This study demonstrates the technical interest of
renewable energy storage. Nevertheless, the economic
justification of such a system must also be explored. In
this regard, the principal indicator is the cost-per-cycle
that represents the storage cost of 1 kWh in a given
technology. To be economically viable, this cost-per-cycle
must be below the gain achievable by the deferred resale
of electricity. Put another way, this cost must be lesser
than the difference between the lowest and the highest
retail price of electricity during the day. According to the
work of Stycynski et al. [23], the only technology that
could be cost-efficient at the moment is the mechanical
storage by pumping (PSPS). Nonetheless, we believe
that R&D investments will drive the storage costs down
and that the variability of wholesale electricity market
prices will increase in order to represent more faithfully
the constraints to which the grid is subjected in reality.
[10] European Power Exchange, Market data day-ahead auction,

Available online at: http://www.epexspot.com/ en/market-data/
dayaheadauction, 2014.
43
Fig. 3: Flywheel simulation model. Modelling of saturation phenomena is not shown to enhance
the figure clarity. Adapted from [21].
[11] I. Gyuk, M. Jonhson, J. Vetrano, K. Lynn, W. Parks, R. Handa, L.

Kannberg, S. Hearne, K. Waldrip, and R. Braccio, Grid energy
storage, tech. rep., U.S. Department of Energy, December 2013.
system applications: a survey, in Electrical Machines (ICEM), 2012

XXth International Conference, pp. 21192125, 2012.
[18] Beacon Power, Flywheel energy storage system, tech. rep., Beacon
Power, 2014.
[12] P. W. Parfomak, Energy storage for power grids and electric

tranportation: A technology assessment, tech. rep., Congressional
Research Service, 2012.
[19] S. Grillo, M. Marinelli, S. Massucco, and F. Silvestro, Optimal

management strategy of a battery-based storage system to improve
renewable energy integration in distribution networks IEEE Trans.
Smart Grid, vol. 3, pp. 950-958, June 2012.
[13] Victron Energy Blue Power, GEL and AGM batteries, Available
online at: http://www.victronenergy.com/upload/documents/
Datasheet-GEL-and-AGM-Batteries-EN.pdf , 2014.
[20] S. Teleke, M. E. Baran, S. Bhattacharya, and A. Q. Huang, Optimal

Control of Battery Energy Storage for Wind Farm Dispatching,
IEEE Trans. Energy Convers., vol. 25, 2010.
[14] CEA, Lhydrogne, les nouvelles technologies de lnergie,

Available online at: http://www.cea.fr/technologies/l-hydrogeneles-nouvelles-technologies-de-l-ene, 2005.
[21] M. Azubalis, A. Baranauskas, and G. Tamulis, Wind power

balancing using flywheel energy storage system, Elektronika Ir
Elektrotechnika, vol. 19, 2013.
[15] G. Roe, Boing flywheel energy storage technology, tech. rep., Boing
Research and Technology, 2012.
[16] R. Pena-Alzola, R. Sebastian, J. Quesada, and A. Colmenar,
Review of flywheel based energy system, in Proceedings of the
2011 International Conference on Power Engineering, Energy and
Electrical Drives, (Torremolinos, Spain), May 2011.
[22] O. Tremblay and L.-A. Dessaint, Experimental validation of a

battery dynamic model for EV applications, World Electric Vehicle
Journal, 2009.
[23] Working group C6.15, Electric energy storage systems, tech. rep.
CIGRE, 2011.
[17] M. I. Daoud, A. S. Abdel-Khalik, A. Massoud, S. Ahmed, and N. H.

Abbasy, On the development of flywheel storage systems for power
Fig. 4: Battery model used to simulate charge and discharge cycles. Based on [22].
44
Fig. 5: Evolution of according to time for flywheels (blue) and batteries (red).
Biography
Pierre-Alexandre Levy is an engineer student at INSA
Lyon, France. He worked at Lawrence Berkeley National
Laboratory on Homogeneous Charge Compression
Ignition (HCCI) combustion for small generators in
2014. For his end-of-study internship, he is employed by
Schneider Electric for automation and regulation in the
Heating, Ventilation and Air Conditioning (HVAC) field.
He will received the Dipl.-Ing. degree from the Energy
and Environmental Department of INSA Lyon in 2015.
Guillaume Mancini is an engineer student at INSA

Lyon, France. His research experiences include turbulent
combustion modelling for Rolls-Royce (2013), and Smart
Grid analysis with the ISS4E team (2014), in Waterloo,
Canada. He is currently working at EDF (lectricit de
France) in the field of photovoltaic systems, electrical
grid and energy storage. He will received the Dipl.-Ing.
degree from the Energy and Environmental Department
of INSA Lyon in 2015.
45
Wind induced motion on bundled conductors

(Excluding ice galloping)
Part A - Aeolian vibrations
Study committee B2
Working group B2.66
Convener: G.Diana, Secretary: H.J.Krispin
Members:
Jennifer Havel, Jason Huang, Julien Garnier, Sergey Kolosov, Jean Philippe Paradis, Andr Leblond ,
Chuck Rawlins, Umberto Cosmai, Pierre Van Dyke, Jean-Louis Lilien, Wolfgang Troppauer,
Alessandra Manenti , Laura Mazzola, E. Ruggeri, Naji Sahlani, Douglas Proctor, Ccile Roz,
Marta Landeira, Sergio Thaddey, Claude Hardy, Ronald Tong, Bin Liu, A.Bhangor, A Vinogradov,
T.Furtado, M.Araujo.
1. Introduction
are reported in [1][2][3][5] to which we make reference for

more details.
Research conducted by Working Group B2.31 (formerly TF

1 of WG B2-11) and summarised in reports [1] [2] & [3],
covers the modelling of Aeolian vibration of an undamped
single conductor and of single conductor plus dampers.
It has been shown that analytical methods based on the
Energy Balance Principle (EBP in the following) and a
shaker-based technology can provide a useful design tool
for damping systems that protect a single conductor against
Aeolian vibration.
However, from the engineering point of view, the EBP is a

very useful tool for the design of damping system necessary
to control Aeolian vibrations.
2. Analytical methods and

assumptions used in the study
The study was carried out by comparing the results of EBP
calculations coming from three different experts. All experts
based their calculations on the same data sets describing the
characteristics of the spacer-dampers, however there were
nominal differences in analytical procedures, such as the
wind power input functions, self-damping models applied
in the EBP calculations, or the modelling of the effects of
wind turbulence. Table I summarizes these similarities and
differences taken as main assumptions in the simulations.
Although the above differences, all the models are based
upon the same assumptions and they are developed in
the frequency domain, hence a preliminary evaluation of
the bundle natural frequencies and modes of vibration is
necessary [1][2]. To do this, for each mode of vibration
the Aeolian vibration amplitudes are achieved through a
balance between the power input from the wind, the power
dissipated by the conductors and the damping devices
(spacers and dampers).
One of the purpose of the present report is to evaluate

the effectiveness of these methods for the design and/or
verification of the damping system of conductor bundle
spans with respect to Aeolian vibrations.
As in the case of the former papers, this report is based on
an analysis of the available technology and on the results
of benchmarks: an analytical-analytical benchmark and
an analytical-experimental one are used for the evaluation
relevant to Aeolian vibrations. The comparison between
the analytical results produced by the different available
models and the experimental one will help to understand
the limitations and the usefulness of considered approach.
Field tests relevant to a 500 m quad bundle span equipped
with ACSR Drake conductor and spacer-dampers have
been selected as test case both for the analytical-analytical
and analytical-experimental benchmark for the Aeolian
vibrations evaluation. The selected case has a H/w
parameter around 2000 m with respect to the safe design
tension requirements according to [4] that is 2500 m.
In all the models the following assumptions are considered:

The vertical and horizontal components of the conductor
motion are taken into account The wind power input is
only related to amplitudes in the vertical plane [6];
The conductor self-damping is due to amplitudes
resulting from vertical and horizontal motion;
The spacer-dampers are modelled through their inertial,
It must be pointed out that EBP based methods do not

simulate the full complexity of the problem. The complexity
of the problem and the limitations of the EBP approach,
together with the more sophisticated tools nowadays
available to reproduce the Aeolian vibrations phenomenon
46
In the following a brief description of the three models

considered is provided.
elastic and damping characteristics, as detailed in Table

III. The coupling among sub-conductors is defined by
the spacer dynamic response, which is a function of the
vibration frequency;
The aerodynamic damping related to the horizontal
motion is not considered;
The torsional stiffness of the conductor is not considered
in the computation of the bundle natural frequencies;
The longitudinal motion is not considered;
The energy input from the wind comes from wind
tunnel tests on two vibrating cylinders, with one in the
wake of the other [5][6][7]:
Half of the overall energy derived from the wind
tunnel tests on two cylinders is applied to any
conductor of the bundle (independently on the subconductors number)
Because the energy is a function of the amplitude of
motion of the conductor, each sub-conductor may
have different wind power input.
Krispin Model
The computational model determines vibration modes
of the conductor bundle [11]. For that purpose, the
subconductors are assumed to behave like strings with
small bending stiffness. The subconductors are divided
into subspans at the locations, where spacers are attached.
The spacers are represented by an impedance matrix which
describes the relation between the conductor forces and
velocity at the spacer clamps. Stockbridge dampers are
treated analogously.
Formulating the equation of motions of the subspans
and taking into account boundary conditions (clamped
span ends) and compatibility conditions at the spacer
clamps, leads to a set of homogeneous equations. Solving
Analytical Methods
Diana et al
Krispin
Cosmai (Claren)
Case data
AA+AE B.
Table II
Table II
Table II
Spacer Characteristics
AA+AE B.
Table III
Table III
Table III
Wind Power Data
AA B.
Curve b Table IV [8]
Curve b Table
IV [8]
AE B.
Table IV [7][8]
[11], [13]
Curve b Table IV [8]
YES [7][8]
NO
YES (Variable turbulence)
AA B.
Table IV
Table IV
AE B.
[10]
[11], [13]
[10]
Calculation Method
[10]
[11], [13]
[10]
Approach
Matrix transfer
method.
Modal approach
Complex Eigenmodes
Matrix transfer method.

Modal approach
Consider Flexural Stiffness
0.5 EJmax
0.5 EJmax
0.5 EJmax
Consider Turbulence
Self-Damping Data
Tensile load differentials
AA B.
No
No
Tensile load differentials
AE B.
Yes
Yes
No
Energy Balance domain
Single Freq
Single Freq
Single Freq
Spacer Model
[10]
[11], [12]
[10]
Possible to Consider Dampers
Yes
Yes
Yes
Possible to Consider Armour rods
Yes
No
Yes
Table I - Analytical Methods and Assumptions Used in the Study (Numerical table entries refer to supporting references)
Note: AA B. stands for Analytical-Analytical Benchmark, while AE B. stands for Analytical-Experimental Benchmark.
47
the eigenvalue problem gives the complex eigenvalues.

The corresponding eigenvectors define the complex
eigenmodes.
The EBP method is employed to evaluate the vibration

amplitude of these modes. Complex mode shapes have
antinode and node amplitudes that vary along the span.
The complex mode shape is approximated by an equivalent
standing wave possessing the same mechanical energy as the
complex wave. Conductor self-damping and wind power
input are evaluated for the amplitudes of this equivalent
standing wave. Power dissipation of the spacers is derived
from the clamp velocities and the spacer impedance.
The maximum amplitude of oscillation over the whole

span;
The maximum amplitude of oscillation registered in
correspondence of spacer and damper clamps over the
whole span
The strain in correspondence of singularities of the
system: i.e at the suspension clamp, at the armour rods
and at the spacer and damper clamps;
It must be pointed out that the matrix transfer method,

applied to the above described approach, has been also
applied to the models developed in Russia for vibration of
single [14,15] and bundle conductors.
Claren - Cosmai Model:
Vibration intensity is calculated in terms of the maximum

value of .max, which is the product of frequency and
antinode amplitude in the respective subspan. Nominal
strains at clamping points are derived by using well-known
relationships between nominal strain and .max.
This model [8] [10] is the same as the Diana model, using
the same software.
However during the years some changes have been
introduced in the formulation of the energy dissipated by
the conductors and of the energy introduced by the wind.
Diana Model:
In this model, eigen frequencies and eigenmodes of
the system (real modes) are computed through the
matrix transfer method. Field matrices [Bi] (i=1,..,N
with N=number of subspans) defining the relationship
between displacements and forces on the conductors at the
extremities of each subspan and point matrices [Pj] (j=1,
.,N-1) defining the relationship between displacement
and forces on the conductors right side and left side of
each spacer are defined. The product of all the field and
point matrices, gives a matrix [A] defining the relationship
between displacements and forces at the span extremities.
End conditions allow for the computation of the eigen
frequencies and correspondent eigen modes [7,8,9,10]
3. Benchmarks
In this section the results obtained for two different
benchmarks are summarised: the first one is an analyticalanalytical benchmark, i.e. the amplitudes and strains
predicted by the various available models applied to a certain
test case are compared; the second one is an analyticalexperimental benchmark: in this case strains predicted by
analysis are compared to strains actually measured in a field
test in Pakistan. The experimental data were available to the
benchmark participants since the beginning. In table II the
main information concerning the considered experimental
tests are reported, while table III summarises the spacer
data The input data of tables II and III are used by both the
benchmarks.
Hence in the Diana model the EBP is applied for each

computed eigenfrequency/eigenmode considering both
the energy introduced by the wind and the one dissipated
by the system.
In order to provide comparable results and highlight the

main differences in the models, the same energy input from
the wind and the same dissipated energy for the conductors
have been introduced in the analytical-analytical benchmark
(see table IV).
All the damping sources, such as spacers dampers and,

cables are accounted for. For the eigenfrequencies and
eigenmode where the amplitude of oscillations are not zero,
the amplitude of motion and the resulting deformations
are computed. Moreover the model allows to account for
frequency dependent spacer stiffness and damping and
tension differential between bundle conductors.
Analytical Analytical benchmark results:

For the analytical-analytical benchmark the two considered
models are the ones of Krispin and Diana, both of them
Finally the main outputs of the model consist in:
48
Characteristics of the Test Span Section 1 (Tarbela-Jalapur-Sharif)

Location
Between towers 443-444 ( susp - susp)
Span length (l) [m]
449
Terrain condition
Broken area, open terrain with low vegetation
Elevation above sea level [m]
494
Number of circuits
Phase configuration
Quad bundle, 457mm separation
Phase conductor
Type and name
ACSR Drake
Diameter (D) [mm]
28.14
Mass per unit length (m) [kg/m]
1.628
Ultimate Tensile Strength (UTS) [N]
139060
Tensile load (T) [N]
32000
Stranding
7 steel wires + 26 aluminum wires
Elementary wire diameters [mm]
3.45 (steel) - 4.44 (aluminum)
Spacer damper
Number
Subspan lengths [m]
37 63 55 68 61 69 59 - 37
Table II Experimental tests characteristics and system configuration data
use the same functions for wind power and conductor selfdamping as shown in Table IV; it is then easier, from the
results, to highlight the structural differences in the models.
The Claren-Cosmai model is not considered for this type
of benchmark, because it should provide exactly the same
results as the Diana model.
The predicted maximum vibration amplitudes for each
mode of vibration are reported in figure A.2 - as a function
of frequency - for the Diana and Krispin model. The
predicted maximum strains are reported in figure A.3.
In both the cases the same tensile load is applied on
the conductors of the bundle and a constant low wind
turbulence is imposed, according to Table I and table IV.
49
Central body mass
2.177
kg
Arm mass
0.735
kg
Central body Moment of inertia
6.47 10-2
kg m2
Arm Moment of inertia
1.57 10-3
kg m2
Hinge torsional stiffness KT
333
Nm/rad
Torsional loss factor
0.35
HT/KT
Hinge axial stiffness KA
100
kN/m
Axial loss factor
0.2
HA/KA
Table III spacer-damper data
Wind Power Input
Conductor Self-damping
W/l = power dissipated per unit length of conductor [W/m]

W/l = DC U2 f4
l = span length [m]
U = vibration amplitude [m]
f = vibration frequency [Hz]
conductor data:
m = mass per unit length = 1.628 kg/m;
T = tensile load = 32000 N;
D = diameter = 0.02814 m
DC =damping constant = 0.003089
Curves b,c,d are wind power input curves on each one of the bundle
conductors, for different turbulence levels.
The wind power input used in the analytical-analytical benchmark is
that represented by curve b,
corresponding to low turbulence (It < 0.07)
Conductor self-damping used for

the analytical-analytical benchmark
Table IV wind power input and conductor self-damping for the analytical-analytical benchmark
As far as the maximum strain is considered, figure A.3

shows the bundle maximum strain both in correspondence
of the spacer clamp (green triangles and red diamonds) and
the suspension clamp (violet triangles and blue diamonds),
for Krispin (triangle points) and Diana models (diamonds
points).
In figure A.2:
- the continuous lines, reported for reference purpose,
show the results obtained in the case of a single conductor
at the same tensile load of the bundle conductors (the
blue line refers to Diana Model, the red one to Krispin
model),
- the blue diamonds (Diana) and red crosses (Krispin)
represent the maximum amplitude observed on the
bundle conductors.
Also in this case, for the whole range of frequency

considered, a general good agreement between the results
can be observed. There is a good agreement between the
two model results for frequencies up to 25 Hz, while at
higher frequencies, where the amplitude of vibration for
both the models are very low (see figure A.2), the Krispin
model shows a level of strain higher than the one of Diana.
The diagram shows that, even if there are some

differences in the two models, the trend in the results
is very similar.
50
Figure A.2- Predicted maximum Aeolian vibration antinode amplitude (0-peak) as a function of frequency when the same
tensile load is applied on the subconductors and a constant low wind turbulence is considered
Analytical Experimental benchmark results
The experimental results for the test case are available in

the form of strains (0-peak value) at the suspension clamp,
deduced by bending amplitude measurement performed
through a typical vibration recorder.
In the analytical-experimental benchmark the models described in section 2 - use for the wind power input and
conductor self-damping their own selected function, as
shown in Table I.
Experimental and numerical results are compared, for the

different models, in figures A.4, A.5, A.6 when the same
tensile load is applied to each sub conductor and a constant
low turbulence is considered. The choice of a low turbulence
level is related to the terrain description given in Table II.
Moreover, the data describing the field span and the spacer
are the same as for the analytical-analytical benchmark
(table II and III).
Figure A.3 - Predicted maximum strain (0-peak) on the bundle conductors at the suspension clamp
and at the spacer clamp as a function of frequency
51
Figure A.4: Maximum strains at the suspension clamp with the Diana model : when the same tensile load is applied to the
subconductors of the bundle and a constant turbulence (It < 0.07) is considered
results are still conservative and similar to the previous

ones.
Figure A.4 shows the maximum strain for the bundle

conductors in the case of the Diana model. It can be
seen that, when the same tensile load is applied on the
subconductors and a constant low wind turbulence index
It lower than 0.07 is considered, the numerical data are
conservative with respect to experiments in the frequency
range 5-20 Hz.
In order to achieve a better understanding of the

discrepancies highlighted before between experimental
and numerical results, a sensitivity analysis (figure A.7,
A.8, A.9, A.10) has been performed, within the benchmark.
Two main factors have been considered: the first is the
wind turbulence, that is generally not constant with the
wind speed, the second is that the tensile load cannot
be exactly the same on all the bundle subconductors.
Its a practice in the bundle design to string the bottom
conductors with a lower tensile load with respect to upper
ones. Hence the modes of vibration will be affected by
this tension differential [12]. The tension differential
is expressed in terms of difference in sag, measured in
conductor diameters (D): sensitivity analysis has been
done considering a 1D and a 10D situation.
In figure A.5 the same analysis described for figure A.4 is

replicated using the Claren-Cosmai model. As expected,
the results appear to be similar to the one obtained by
means the Diana Model
The two considered turbulence levels (It < 0.07 and It
= 0.05) have to be considered equivalent, representing
the lowest turbulence levels present in each of the two
computation programs.
Finally, the same analyses as before are replicated by means
of the Krispin model. As it can be seen, the numerical
As a first step the influence in the results of variable wind
Figure A.5: Maximum strains at suspension clamp with Claren-Cosmai model : when the same tensile load is applied to the
subconductors of the bundle and a constant turbulence (It = 0.05) is considered
52
Figure A.6: Maximum strains with Krispin model : when the same tensile load is applied to the subconductors
of the bundle and a constant, low, wind turbulence is considered
leads to a strain decrement in the bundle: it can be seen that

the decrease of strain level when a 1D tension differential is
applied is low.
turbulence is considered: a typical trend (black points) of

measured turbulence as function of wind speed is reported
in figure A.71: the continuous line represents the index of
turbulence as function of speed considered in the ClarenCosmai model.
Finally, in figure A.10 the maximum strains at suspension

clamps computed considering a tension differential of
10D is reported. Moreover in this calculation the variation
of stiffness and damping with respect to frequency for
the spacer has been considered. It can be observed that in
order to achieve level of strain close to the experimental
one, a tension differential of 10 D is required. This tension
differential expressed in conductor diameters (10 D)
corresponds to a percentage tension differential between
upper and lower sub-conductors of the order of 2%.
Figure A.8 reports the numerical maximum strains at

the suspension clamp when the above described variable
turbulence is considered (figure A.7): it is evident that the
maximum strains undergo a reduction of around 75% with
respect to the ones presented in figure A.5, and show a good
agreement with the experimental findings.
The influence of tension differential has been also
investigated, using the Diana model (see figure A.9 and
A.10).
4. Conclusions and future work
In figure A.9 the maximum strains at the suspension clamp

are reported when a 1D tension differential is applied to
the bundle. The application of a tension differential always
The two benchmarks assessed within the Working Group

allowed to understand the main differences among the
Figure A.7 Cosmai hypothesis for variable wind turbulence index

- The measured turbulence index (black points) in fig A.7 is not that measured in the site to which the field measurements of the AE Benchmark refer. It only represents
a typical trend of turbulence index, not related to a very flat terrain, showing that at low wind speeds, generally, turbulence increases as the wind speed decreases.
53
Figure A.8: Maximum strains at suspension clamp with the Claren-Cosmai model : tension differential
neglected and variable turbulence (0.20 < It < 0.05) considered
Figure A.9 Maximum strains at suspension clamp with the Diana model : tension differential 1D
and constant turbulence (It < 0.07) considered
Figure A.10: Maximum strains at suspension clamp with Diana model : tension differential 10 D
and constant turbulence (It < 0.07) considered
54
models presently adopted in the field of cable vibrations to

understand and prevent Aeolian vibration.
conductors strung at relatively high tensile load. Application to HV

& UHV lines, Electra 2011 - n.256
[4.] CIGRE TF B2.11.04, Technical Brochure 273 2005, Overhead
conductor safe design tension with respect to Aeolian vibrations
The analytical analytical benchmark showed that the

computed vibration have a very similar trend even if some
differences in the models are present.
[5.] G. Diana, M. Falco, M. Gasparetto , On vibrations due to vortex

shedding induced on two cylinders with one in the wake of the
other. Meccanica, n.3, 1976
As far as the experimental-numerical benchmark, the

numerical results generally exceed the experimental ones
and then they are conservative, at least at low frequencies.
[6.] D. Brika, A. Laneville, The Power Imparted by Wind to a Flexible

Conductor in the Wake of Another, IEEE Power Engineering
Review, Vol. 17, No. 1, 1997.
The benchmark has been developed on one case only,

considering a quadruple bundle: this is of course a limitation
and future work comparing analytical experimental
results from different cases (different bundle configurations
and different conductors) should be planned.
[7.] Belloli, M., F. Resta, D. Rocchi, and A. Zasso, Wind Tunnel

Investigation on Aeroelastic Behaviour of Rigidly Coupled
Cylinders. Proc. 5th Intl Symposium on Cable Dynamics. Santa
Margarita Liguria (Italy). September 15-18 2003
[8.] G. Diana, M. Gasparetto, F. Tavano, U. Cosmai, Field
measurement and field data processing on conductor vibration
(comparison between experimental and analytical results).
CIGRE International Conference on Large High Voltage Electric
Systems, 1982 Session
In any case, it is needed to point out that, generally, when

dealing with twin bundles, numerical results appear to be
less conservative in respect to the experimental data [4][7].
[9.] G. Diana, G. Di Giacomo, R. Claren , An approach to vortex

shedding under turbolent air flow on a single vibrating cylinder.
IEEE 1979 Summer Meeting, July 15, N.Y
The sensitivity analysis demonstrated that a non-negligible

influence in the assessment of cable behaviour, when dealing
with Aeolian vibrations, is given by the introduction of
tension differentials and variable wind turbulence with
wind speed.
[10.] R. Claren, G. Diana, F. Giordana, E. Massa , The vibrations of

transmission line conductor bundles. 71 TP 158 PWR - IEEE
Winter Meeting, 1971
Clearly it is not straightforward knowing the real value to

assign to the turbulence and to the tension differentials
when the bundle behaviour for Aeolian vibrations must be
analysed.
[11.] K. Anderson, P. Hagedorn , On the energy dissipation in spacerdampers in bundled conductor of overhead transmission lines.
Journal of Sound and Vibration, 180(4), pp. 539-556, 19Verma,
H. - Aerodynamic and structural modeling for vortex-excited
vibrations in bundled conductors. Doctors thesis, Darmstadt,
2008
Future work to achieve better knowledge on this issue

should consider comparison between measurements on a
real line and analytical results.
[12.] Foata, M., Noiseux, D., Computer Analysis of Asymmetries of a

Two-Conductor Bundle upon its Aeolian Vibration, IEEE Trans.
on Power Delivery, Vol. 6, No. 3, July 1991
[13.] T. Hadulla, Vortex-induced vibrations in overhead line bundles.
Doctoral Thesis, Technische Universitt Darmstadt, 2000
5. References
[14.] O. Afanasyeva, S. Ryzhov, V. Feldshteyn Dynamic models for

the study of vibration Power Line conductors and communication
cables in the air flow Problems of Mechanical Engineering and
Automation, 1998, No 1, pp. 50 57
[1.] CIGRE TF 22,11,01, Modelling of Aeolian vibration of Single

Conductors - Assessment of Technology. Electra 1998 N.181
[2.] CIGRE TF B2,11,01, Modelling of Aeolian vibration of a single
conductor plus damper: assessment of technology. Electra 2005 - N.223
[15.] O. Afanasyeva, V. Feldshteyn, S. Ryzhov, Calculation of rational

geometry of the helical support clamp under the terms of vibration,
Electricheskie Stantsii (Power Plants) 1998. No 1, pp. 12 17
[3.] CIGRE WG B2.31, Modelling of Aeolian vibration of single
55
Wind induced motion on bundled conductors

(Excluding ice galloping)
Part B Subspan oscillations
Study committee B2
Working group B2.66
Convener: G.Diana, Secretary: H.J.Krispin
Members:
Jennifer Havel, Jason Huang, Julien Garnier, Sergey Kolosov, Jean Philippe Paradis, Andr Leblond ,
Chuck Rawlins, Umberto Cosmai, Pierre Van Dyke, Jean-Louis Lilien, Wolfgang Troppauer,
Alessandra Manenti , Laura Mazzola, E. Ruggeri, Naji Sahlani, Douglas Proctor, Ccile Roz,
Marta Landeira, Sergio Thaddey, Claude Hardy, Ronald Tong, Bin Liu, A.Bhangor, A Vinogradov,
T.Furtado, M.Araujo.
1. Introduction
for the application of the aerodynamic forces to each

subconductor using the QST with a non linear approach.
Subspan oscillations is a well-known phenomenon in High

Voltage and Ultra High Voltage Overhead Transmission
Lines (HV and UHV OHTL) [1]. It occurs on conductor
bundles and it is due to the effect of the wake produced
by the windward conductor on the leeward one. For this
reason the phenomenon is also classified as wake induced
oscillations, this phenomenon is a flutter type instability
due to the coupling of vertical, horizontal modes in a
frequency range between 0.5 2 Hz.
However, FEM analyses in the time domain are not always

a practical tool for subspan oscillation simulation also
because of the computation time required to obtain results.
All the models developed up to now rely on the Quasi
Steady Theory (QST): according to this, the field of forces
acting on the conductors in the wake is identified using
the static aerodynamic coefficients measured in the
wind tunnel and the effect of relative motion between
sub-conductors is accounted for introducing a relative
velocity with respect to the approaching flow [7][10].
This approach generally holds for very high reduced
velocities Vr, being Vr defined as the ratio between the
wind speed V and the product of oscillation frequency f
and cylinder separation I: Vr=V/(f I). Moreover, being the
problem highly nonlinear, the validity of this theory needs
to be confirmed. Another important issue is the Reynolds
number (Re) effect.
Recently, problems associated with this phenomenon

has become more recurrent, attracting the attention of
transmission line operators, hence WG B2.46 decided to
dedicate an effort to evaluate the present developments in
such a field.
As for the analytical models developed for studying such a
problem, the first ones are those of Simpson [2], Ikegami
[3], Diana [4], Ko [5] and Tsui [6]. These are two degrees
of freedom models: the motion of the leeward cylinder is
studied along two orthogonal directions, the windward
cylinder being still. The linearized Quasi Steady Theory
(in the following QST) is employed and the drag and lift
coefficients on the leeward cylinder are deduced from
static measurements in wind tunnel, as a function of the
relative position of the leeward cylinder with respect to the
windward one. Such models are linear and clearly simplify
the structural behaviour of the bundle sub-conductors,
taking it back to a two degrees of freedom system in which
the leeward conductor is the only one moving. Rawlins in
[16][17][18] expresses bundle dynamics in terms of normal
propagation modes using the transfer matrix method. The
bundle stability is analyzed reproducing the aerodynamic
forces through the QST with a linear approach. Nowadays
the Finite Element Model (in the following FEM) analysis
allows for the reproduction of the bundle dynamics and
In fact for stranded conductors, i.e. rough cylinders, with

the typical values of conductor diameter and wind speed
involved by subspan oscillations, Re may be close to the
critical zone: hence the Re number could significantly
affect the phenomenon, due to the non-negligible
variations of the drag coefficient with Re itself [2][8]. As
a matter of fact, recent researches made at Polimi [9] have
shown that the QST maintains its validity for these values
of reduced velocity.
However, the same research put in evidence that
the Reynolds number is very important because the
conductors during subspan oscillations can operate
around the critical range of the drag coefficient (see
figure B.1 [8], where the drag coefficient of a 40.7 mm
56
CHUKAR CONDUCTOR
SMOOTH CYLINDER
1.3
1.2
1.1
1.0
0.9
0.8
0.7
6
1.7
12
3.3
18
5.0
24
6.7
30
8.4
36
10.0
Vrm/s1
Rex10-4
Figure B.1: Drag coefficient on windward rough cylinder as function of speed/

Reynolds number from Wind tunnel experimental tests [8]
Figure B.2: Rough cylinders energy with respect to x/D amplitude for the three considered frequencies: experimental numerical comparison.
Due to the model scale, f = 0.5 Hz corresponds to f = 1 Hz full scale
is both the static and dynamic bundle rotation angle with respect to the wind.
vertical scales are different in the two figures).
diameter Chukar conductor is reported, compared to

that of a smooth cylinder with the same diameter). For
this reason the static tests in wind tunnel to identify the
aerodynamic coefficients on the leeward conductor must
be made on rough cylinders, covering the critical and
supercritical range of the drag coefficient. Such an aspect
has a direct impact on the evaluation of the bundle stability
as demonstrated in Figure B.2, where the measured and
computed energy on a couple of cylinders with one
in the wake of the other is reported for different wind
speeds, when the cylinders are moving along elliptical
orbits [9]. The energy input from the wind reported in
Figure B.2 is a function of the subspan elliptical motion
amplitude, represented on the horizontal axis in nondimensional terms, as ratio x/D between the amplitude
and the conductor diameter.
Additionally, it can be observed that the higher the

frequency of the motion, the more stable the bundle
appears, hence it is confirmed that a benefic effect on
subspan oscillation can be obtained by decreasing the
subspan length.
In the present work a benchmark within the different
type of models at disposal for subspan oscillation
studies is carried out comparing numerical results with
measurements on the IREQ Varennes test span equipped
with a quad bundle of ACSR Bersimis conductors and
spacer-dampers.
2. Analytical methods and

assumptions used in the study
As can be seen, increasing the vibration amplitude, the

wind energy input is decreasing, becoming negative, so
the subspan phenomenon is self-controlled. The Reynolds
effect makes the bundle less stable at low speed, as can be
seen comparing the energy input at 10 m/s to that at 15
m/s in Figure B.2 (please note that, to held readability, the
In this section a brief description of the analytical

methods adopted in the study is provided, together with
the assumptions introduced in the analysis.
57
2.1.Diana Model:
Once the aerodynamic forces on the conductors are

known, it is possible to compute, in each section of the
bundle, the energy introduced by the forces themselves in
one complete elliptical cycle.
The Diana Model is an energy based method that firstly

evaluates the natural frequencies and vibration modes
of the bundles, then identifies the modes showing a
predominant horizontal and vertical component and
selects those in the frequency range 0.5 3 Hz, frequency
range typical of subspan. Subsequently within the
selected modes, the ones which could be coupled by the
aerodynamic forces to give rise to subspan oscillations are
chosen.
The method gives the energy introduced by the wind in

all the bundle, summing up the energies computed in each
section: the evaluation is performed both for + /2 and
/2 phase shift (see equation (1)), obviously choosing the
situation for which the aerodynamic forces can introduce
maximum energy into the bundle system.
Finally, the steady state amplitudes of oscillation are defined
through the balance between the energy introduced by
the aerodynamic forces and the energy dissipated by the
bundle, equipped with the spacer-dampers.
For each possible pair of horizontal and vertical modes,

the possibility of having instability is evaluated by a Quasi
Steady Theory (QST) linear approach and the range
of wind speeds for which subspan oscillation may be
excited together with an instability index are computed.
This preliminary analysis provides also the maximum
difference between the vertical mode frequency and the
horizontal mode frequency allowing for the coupling
between the two modes.
The Reynolds effect plays an important role in the drag

and lift aerodynamic coefficients of the conductors, both
for the windward and the leeward ones. This effect is
introduced in the model thanks to the results of a recent
research conducted in the Wind Tunnel of Polimi [9].
Once the types of modes that can be coupled are

defined, two independent modal coordinates qo(t) for
the horizontal mode and qv(t) for the vertical mode of
the bundle are chosen. Harmonic laws are imposed to
the two coordinates qo(t) and qv(t), in such a way to
reproduce the elliptical motion of the conductors typical
of flutter instability. Being the maximum instability
index found for a phase shift between the two modes of
/2 the imposed law is:
2.2.1. I.Sergey / Vinogradov Model:

The I.Sergey/Vinogradov method [14, 15] relies on the
direct numerical simulations of the conductors and
spacer-dampers movement in the process of subspan
vibrations. It is fully based on the finite differences method
and the explicit scheme of system equations solution. The
equations of movement initially compiled in 3D form, for
the sake of simplicity are reduced to 2D problem taking
into account negligible longitudinal displacements of
Spacer-Dampers along the span length. The calculation
of the bending stresses in the conductors is performed
assuming that the Spacer-Damper clamp divides
conductors bending angles uniformly. The aerodynamic
coefficients are taken after G.Diana [4], S.Price [11].
The assumption that Poffenberger-Swart formula is valid
while subspan oscillation process is taking place is made.
Moreover the model has been demonstrated to be valid
for quad bundle span, where in order to replicate properly
the subspan oscillation an initial condition in terms of
force field is applied on the bundle: two short startup
impulses of the forces acting in opposite directions are
applied on a pair of interacting conductors. In case of even
number (e.g. four) of conductors in the bundle, it turned
out most effective to take into interaction only one pair
of conductors. The solutions taken prevent the model to
find only the snaking motion and allow to detect subspan
(1)
where Sam and Sim represent the maximum amplitudes

for the horizontal and vertical modes, being the modes
normalized in such a way that the maximum amplitude
of the cylinder motion along the whole span is equal to
unit, and is the chosen as the circular frequency of the
horizontal mode. Finally, the conductors motion all along
the span is obtained multiplying qo(t) and qv(t) by the
modal shapes.
Using the QST, the forces acting on the conductors in the
different positions occupied along the elliptical trajectories
are computed, taking into account the velocity of each
conductor [7][10].
58
Spacer-damper data
Quad spacer-damper interaxis
457mm
Torsional stiffness of the hinge
125 Nm/rad
Radial stiffness of the hinge
50000 N/m
Ratio between torsional damping and torsional stiffness
0.35
Ratio between radial damping and radial stiffness:
0.35
Arm mass
0.71 kg
Arm mass moment of inertia with respect to the centre of mass
2.74E-3 kg m2
Central body mass
3.44 kg
Central body mass moment of inertia with respect to the centre of mass:
38.5E-3 kg m2
Mass of spacer damper
6.28 kg
Span data
Conductor:
ACSR Bersimis
diameter
35.1mm
mass
2.185 kg/m
Tensile load
34420 N
Span length
450m
Sub-Span arrangement [m]
40-53-57-50-55-49-58-52-36
Table B.1: Subspan oscillation test case data
The first two load steps for the pretension of the conductor
and the dead load are executed without time integration.
Afterwards a do-loop with time integration and update of
the aerodynamic forces is performed. For the numerical
solution the HHT (Hilber-Hughes-Taylor) algorithm is
applied.
oscillations. But lack of wind energy influx provided lower

sensitivity of the model to weak and moderate winds; this
disadvantage was corrected by a coefficient found with the
help of the benchmark test itself.
2.3. Snegovskiy / Lilien Model :
The Quasi Steady Theory (QST) is used for the computation

of the aerodynamic forces. The resultant polynomials of
the wind tunnel test according [7] are the input for the
model approach. The latest wind tunnel test results [9]
are also included, which quantify the influence of the
Reynolds number onto the aerodynamic coefficients.
The corresponding model relies on the use of finiteelement nonlinear formulation. Within this approach, the
interaction of subconductors due to the wake is represented
using Simpsons aeroelastic model [2]. A special force
element is created to introduce the aerodynamic loads
due to the wake which are computed according to the
QST. Moreover the aeroelastic properties of the wake
force field are tuned to meet the wake-induced instability
properties, as measured by Price [11]. Extension of the
wake interaction sample onto the full line span (including
spacer-damper, any bundle configuration) is done taking
into account the inertia-stiffness properties of the line
fittings (spacer dampers). More details on [12], [13].
3. Experimental and analytical

benchmark : description and
results
It must be pointed out from the first beginning that the
models using FEM can give rise, at wind velocities between
10 20 m/s, as reported in detail in [12], to an instability
motion of the whole span at very low frequency also called
galloping or snaking motion. This instability motion is
predominant and this approach is generally not able to
reproduce subspan oscillation. However a full description
of the results obtained by FEM model is reported in [12].
2.4 Kurmann Model :

The idea of the model approach is a finite element implicit
transient analysis with an update of the aerodynamic
forces after each time step. The geometry model consists
of linear beam elements for the conductor, the frame and
the arms of the spacer damper. The stiffness and damping
properties of the spacer-damper hinges are considered
as spring-damper elements. The self-damping of the
conductor is implemented with the Rayleigh damping
model.
The I.Sergey-Vinogradov model, on the other hand, in

order to excite subspan uses as initial condition an external
excitation to move the subconductors out of phase.
59
Figure B.3: Test set-up showing towers, conductors and anemometers of the test line
The subspan oscillations were measured in the middle

of each subspan on the bottom North-East conductor
to obtain the horizontal component of the first subspan
mode antinode amplitude. The data was measured during
four weeks every ten minutes at a rate of 32.3pts/s for a
duration of 145.5s. The time signal was then processed
to determine the peak-to-peak amplitude of the most
severe vibration cycle during the recording period and the
vibration frequency. Only the recordings with an apparent
frequency corresponding to the first subspan mode of
the subspan considered and its adjacent subspan were
selected. The purpose here was to distinguish between
subspan oscillations appearing in the above-mentioned
frequency range from snaking oscillations, which usually
occur at lower frequencies, and also from rain vibrations
or aeolian vibrations, which have higher frequencies.
By means of this approach the model is able to reproduce

subspan oscillation in a reasonably short time (after 30s of
real time process development).
In order to compare the models accuracy and reliability an
experimental-numerical benchmark has been carried out,
considering as a reference case the one tested in Varennes
by IREQ, which is described in table B.1.
The compared quantities are the maximum peak to peak
oscillations of the subconductors in the most critical
sections of the span, during subspan oscillations.
The experimental benchmark was obtained on the HydroQuebec full scale test line which has two dead-end spans
and three suspension spans. The length of the spans is
successively 150, 400, 450, 425, and 150m. The subspan
oscillations were measured on the middle span (450-m
span).
Figure B.4 reports the maximum peak to peak conductor

amplitudes measured in the different subspans.
Figure B.4: IREQ Measurements: Maximum peak to peak horizontal oscillations as function of mean wind speed.
60
Figure B.5: Benchmark results: Maximum peak to peak horizontal oscillations as a function of mean wind speed.
Subspan Oscillations is a complex vibration on bundle

conductors. It needs relatively high winds. Important
parameters are :
The numerical simulations have been carried out in the

speed range 20 60 km/h.
Figure B.5 shows, as a function of mean wind speed, the
peak to peak horizontal oscillation amplitude registered in
the subspan 7, which is found to be the most critical one.
Bundle tilt
Ratio vertical to horizontal frequency in each
subspan Tensile load
Ratio between bundle separation I and conductor
diameter D (I/D)
Subspan length
Simple and advanced methods can be used nowadays.
Complex FEM model have been applied with success but
they need a very cumbersome analysis which is very much
depending on some details.
In blue the results from the Diana model are reported,

in green the results from the I.Sergey-Vinogradov model
and in black the results from the [12.] Snegovskiy /
Lilien model, while in red the experimental points are
represented.
It is possible to observe that the Diana and I.SergeyVinogradov numerical models correctly replicate the trend
of the oscillation peak to peak values, which increase with
the increase of mean wind speed up to reach a maximum
in correspondence of a mean wind speed of 35 km/h, then
the oscillation amplitude starts to decrease.
Methods based on modal analysis and energy approach

seem a more useful tool for practical applications.
Moreover the presented results, even if applied to one only
test case, allow to state that:
The numerically computed amplitudes are also in good

agreement with the ones experimentally registered.
4. Conclusions
The study presented several approaches to the evaluation

of the subspan phenomenon, ranging from approaches
based on the EBP to approaches relying on FEM modelling.
The quasi-steady theory seems able to well reproduce

the aerodynamic forces produced during subspan
oscillations;
The Reynolds number affects in a large amount the
energy introduced by the wind. In fact the flow on the
conductors can vary from sub-critical critical and
super-critical region of Reynolds number depending
on the conductor diameter, surface roughness
61
and, finally, wind speed. In the critical region the

energy introduced is greater due to the decrease of
aerodynamic damping on the upstream cylinder
associated to the negative slope of the drag coefficient.
It can be observed that, considering two conductors
with the same diameter but different surface roughness,
the critical region moves towards higher wind speeds
for smoother conductors, like those with trapezoidal
wires - with respect to the standard conductors with
circular strands. This aspect of the problem is for sure
worth of further investigation through suitable wind
tunnel tests.
The numerical model based on EBP approach and
on sophisticated Wind Tunnel tests to identify the
aerodynamic parameters seems a useful tool for
analyzing the subspan oscillations phenomenon, as
shown by the comparison between its results and
the measurements on a full scale transmission line
structure subjected to the real wind.
The numerical model based on the finite differences
method and the explicit scheme of system equations
solution, even if it had to be calibrated with the help
of the benchmark test itself, also seems in condition
to well reproduce the phenomenon. However more
results and analyses are required to confirm its validity.
[6.]
[7.]
[8.]
[9.]
[10.]
[11.]
[12.]
From the analyses performed in this research and, more

precisely, from the results shown in figure B.2, it is possible
to conclude that one way of controlling subspan oscillations
is to increase the subspan oscillation frequency decreasing
the subspan length.
[13.]
[14]
5. References
[1.]
[2.]
[3.]
[4.]
[5.]
[15]
EPRI (2009) - Transmission Line Reference Book WindInduced Conductor Motion: Second Edition, Electric Power
Research Institute, Palo Alto, Ca 2009 1018554, USA page 3-25
Simpson A (1971) On the flutter of a smooth circular cylinder
in a wake - The aeronautical Quarterly
Ikegami R et al. (1971) Structural analysis. High Voltage
Power Transmission Systems Boeing Co. Seattle BPA Ctr
14-03-1362N.
Diana G, Giordana F (1972) Sulle vibrazioni di un cilindro in
scia di un altro Analisi del fenomeno con modello analitico
LEnergia Elettrica (1972) No.7 pp 448-57
Ko R G (1973) Theoretical investigation for Hydro-Quebec into
the aerodynamic stability of bundled power line conductors, Part I:
two dimensional stability analysis of a conductor in the wake of a
fixed conductor NRC (Canada) Lan Tech Rep LA-122
[16]
[17]
[18]
Tsui Y T (1975) Two dimensional stability analysis of a

circular conductor in the wake of another IEEE Paper A75
576-9. [9.]
Diana G, Gasparetto M (1972) Energy method for computing
the amplitude of vibration of conductor bundles due to wake
effect LEnergia Elettrica
R. L. Wardlaw and K. R. Cooper (1973) A wind tunnel
investigation of the steady aerodynamic forces on smooth and
stranded twin bundled power conductors for the Aluminum
Company of America, Lab. Tech. Rep. LA-117, NRC (Canada)
Diana G, Belloli M, Giappino S, Manenti A, Mazzola L, Muggiasca
S, Zuin A (2013) - Wind Tunnel testing developments in the
last 50 years 2013 European and African Conference on Wind
Engineering - Robinson College, Cambridge, UK, July 7-11, 2013
Diana G, Gasparetto M, Di Giacomo G, Nicolini P (1974) Analytical method for computing subspan oscillation. Analytical
and Experimental results - C 74 493-3, IEEE Summer Meeting,
July 14-19, 1974 Anaheim California, USA
Price, S.J., Padoussis, M.P. (1984) - The aerodynamic forces
acting on groups of two and three circular cylinders when
subject to a cross-flow, J. of Wind Engineering and Industrial
Aerodynamics, 17, 329-347 (1984/2)
Snegovskiy, D. & Lilien, J.L. (2010) - Nonlinear finite element
approach to simulate wake-induced oscillation in transmission
system; Proceedings of ASME 2010 3rd Joint US-european
Fluids Engineering (FEDSM2010); available on http://hdl.
handle.net/2268/102157
Snegovskiy, D. (2010) - Wake-Induced Oscillations
in
Cable
Structures:
Finite
Element
Approach,
available
on
http://bictel.ulg.ac.be/ETD-db/collection/
available/ULgetd-07142010-102207/
M.I.Streliuk, I.I.Serguey, A.A.Vinogradov, V.A.Krasnov, (1986)
Dynamic loads on spacers at subspan oscillations of the EHV
T/L conductors, Energ. Stroit., 1986, No 1, pp.67-71 (in Russian)
Sergey I., Vinogradov A.(2003), Mathematical simulation
and calculation experiment for different forms of conductors
dynamics, Proc 5th International Symposium on Cable
dynamics, Santa Margherita, Italy - 15-18 Sept 2003. pp.93-100
C. B. Rawlins, Effect of Wind Turbulence in Wake-Induced
Oscillations of Bundled Conductors, Conference Paper C74
444-6, IEEE PES Summer Meeting, Anaheim, CA, July 14-19,
1974.
C. B. Rawlins,Fundamental Concepts in the Analysis of
Wake-Induced Oscillation of Bundled Conductors, IEEE,
Transactions on Power Apparatus & Systems, Vol. PAS-95, No.
4, July/August 1976, pp 1377-93.
C. B. Rawlins,Extended Analysis of Wake-Induced Oscillation
of Bundled Conductors, IEEE Transactions on Power Apparatus
& Systems, Vol. PAS-96, no. 5, Sept/Oct 1977, pp 1681-89.
62
Evaluation of High Surge Impedance

Loading (HSIL) solutions for increased
natural capacity of 500 kV Overhead Lines.
Study Committee B2
Working Group WG B2.38
Members
Angel Gallego Del Monte, Danna Liebhaber, Hee-Sung Ahn, Jarlath Doyle,
Jos Manuel Frenandez Davila, Oswaldo Junior Regis, Paul Penserini,
Tsinghua University, Vivendhra Naidoo.
Abstract
The third generation of 500 kV lines were developed

after several studies of High Surge Impedance Loading
- HSIL, for a new level of 1200MW, applying the concept
of compaction of the distance between phases, or the
concept of expanded bundles, or a mix of both concepts.
This report compares six different 500kV Transmission

Lines Projects, each with four subconductors per phase
and Surge Impedance Loading (natural capacity) of 1200
MW, which are currently operating in more than 10.000
km of extension in Brazil. Transmission Line compaction
and the use of expanded bundle techniques are analyzed
focusing on electrical coupling and its influence on
positive sequence and zero sequence impedance. The
electric field on the conductors surface and the electric
and magnetic fields at ground level are evaluated for
each configuration.
2. Theoretical aspects
The basic theory of optimization [1] of HSIL shows
that, for the same voltage level, the equalization and
maximization of the electric field on the sub-conductors
surface, and/or an increase of the number of subconductors per phase, increases the Surge Impedance
Loading SIL of the lines.
1. Introduction
Considering the impedance theory, it is also possible to

show that the Natural Capacity of transmission increases
as the line phases are compacted [2] (i.e. the distance
between the phases is reduced), and as the bundle
dimensions are expanded [3] (i.e. the distance between
sub conductors of a phase is increased).
The Surge Impedance Loading (SIL), also called Natural

Capacity, of an alternating current (AC) transmission
line can be defined as the active power that, being
transported through the line, causes a consumption of
reactive power equal to the value of the reactive power
generated by the capacitance of the same line. Power
flows below the SIL produce minimum voltage drop and
stability problems on the line.
2.1 Diagram of a transmission line

The diagram below represents a transmission line in a
system study, where P is the transmitted power and Vs
and Vr are the sending and receiving end voltages. The
parameters of the line are: C (the shunt capacitance) that
provides reactive power and Z (the series impedance).
The series inductance X consumes reactive power and
is responsible for the drop of the voltage and for stability
problems on the long distance transmission.
The lines thermal capacity, which concerns premature

aging of conductors and connectors as well as maintaining
minimum electrical safety clearances, is generally higher
than the natural capacity and is not covered here.
The systems analysis view is especially important for
long distance interconnections where increased Surge
Impedance Loading (SIL) of the lines can reduce or even
remove restrictions on power flow.
In Brazil, the first generation of 500kV lines had a
3-conductor bundle per phase and a SIL of 900 MW.
The second generation of 500 kV lines had 4-conductor
bundles per phase and a SIL of 1000 MW. The need for
long lines, mainly interconnections between the north,
northeast and southeast regions, has motivated the
development of lines with higher transmission capacity.
Diagram : Schematic representation of a Transmission

Line in electric systems studies
63
Figure 1 Racket Tower; the drawings distances and heights are in meters
Figure 2 Cross-Rope Tower (Chainette)
Figure 3 VX-Asymmetrical Tower
self impedance and the positive sequence impedance, Zl,

increasing the capacity of the transmission line.
For steady state and transient studies, the TL is

represented by its positive and zero sequence constants.
In the load-flow studies that analyze the voltage control
in the system, only the positive sequence constants are
used. Therefore, a line that has smaller positive sequence
impedance per unit length (Z1) has higher capacity of
transmission for the same length reference.
The mutual impedance (Zm) depends on the distance

between the phases. As the distance between the phases
decreases, the higher will be the lines mutual impedance.
Therefore, the impedance Z1 will again be smaller cause
Zm has a negative signal in relation (Z1 = Zs - Zm),
increasing the capacity of power transmission.
2.2 Analisys of self and mutual impedance, and zero

and positive sequence
These two techniques, compaction of phase spacing

and expansion of the bundles, were used in different
projects to increase the Natural Power (SIL) of 500 kV
interconnection lines in Brazil, as shown in the next
topic.
The positive sequence impedance Z1 of a line can be

obtained from their self and mutual impedances (Zs and
Zm) by the relation Z1= Zs-Zm.
The self impedance (Zs) depends on the conductor, but
mainly depends on the geometry of the bundle of each
phase. The higher the bundle dimension, the smaller the
While the reduction in phase-phase spacing and the

increase in bundle diameter increases the Natural Power
64
Figure 4 VX-Symmetrical Tower
Figure 5 Cat Face Tower
Figure 6 Monopole tower
Distances and heights are in meters. In all cases, the phase

bundles consist of 4 Rail, 954MCM sub-conductors, but
with different bundle diameters and different distances
between the phases, in order to obtain a Natural Power
of 1200MW.
on the other hand, the influence of those variations of

the self and mutual impedances (Zs and Zm) in the
impedance of zero sequence (Zo) shows a different effect
as the relation is Zo=Zs+2xZm.
Differently from de effect on Z1, the application of
compaction or expansion techniques of the bundles
results in variations of the impedance Zo in both ways,
increasing or decreasing , depending on the case. The
following item (7) shows the Table II with the parameters
of sequence and the analysis of their variations.
3.1 Racket tower

The Racket Tower is aself-supporting structure with
square conventional bundles, 18 inches (0,457m)
on each side. The distance between the phases is very
reduced (bigger Zm) in a triangular arrangement called
Compact Tower. This leads to an increased SIL.
3. Towers and arrangements

analyzed
3.2 Cross-rope tower (chainette)
Pictures of Towers and drawings of phases bundles

showing their sub-conductors are presented below.
The Cross-Rope Tower is supported by two guyed poles,

with the electric phases between them. The bundles are
65
Figure 7 Electric superficial gradients in 500kV
guyed V supports, with horizontal phases having

structure masts between them. In order to obtain a
higher SIL, a regular square expanded bundle is used
(lower Zs), 1,20m/side. This allows the use of standard
hardware which the previous design did not. On the
other hand, as it is shown later, the electric field is higher
in some sub-conductors, since it is not the optimized
electric distribution.
conventional (square with 0,457 meters), with a nearly flat

arrangement of phases and with very short distances between
them, which gives a higher SIL due to the compaction.
3.3 VX-ASYMMETRICAL (V guyed tower with
Asymmetrical Expanded Bundle)
The VX-Asymmetrical Tower has a guyed V support
for the center phase, with phases on the same level
and structure masts between them, which prevent
compaction. To obtain a higher SIL an asymmetrical
expanded bundle (lower Zs) which optimizes the electric
fields on each sub-conductors surface.
3.5 Cat face tower

The Cat Face Tower is self-supported with metal parts
between phases. This limits its compaction. In order
to obtain higher SIL, a regular expanded bundle was
used (lower Zs) with a square form, 1,20m/side. The
phases are arranged in a triangle. This provides a
more equalized distribution of electric fields in the
sub-conductors.
3.4 VX-SYMMETRICAL (V guyed tower with

Symmetrical Expanded Bundle)
The VX-Symmetric Tower has also its conception in
Electric field
kV/cm
Cross Rope
Racket
VX Symmetric
Monopole
VX asymmetric
Cat face
Maximum value
18.00
17.76
17.02
16.88
16.05
16.36
Minimum value
15.04
15.00
14.69
14.52
14.67
14.47
Max / Min (%)
19.7
18.4
15.9
16.3
9.4
13.1
Average - phase A
15.7
15.7
15.3
15.1
15.1
15.0
Average - phase B
17.9
17.3
17.0
16.5
16.0
16.0
Average - phase C
15.7
15.7
15.3
15.1
15.1
15.0
Average - General
16.4
16.2
15.9
15.6
15.4
15.3
Table I: Minimum, Maximum and Average values of electric field on conductors surface.
66
Figure 8 Graphic with the audible noise profiles (AN)
Figure 9 Graphic with radio interference profiles (RI)
3.6 Monopole tower
The Table I below presents, for each configuration,

the values of maximum and minimum gradients, and
the ratio of these values. It also presents the average
of the gradients on sub-conductors of each phase, and
the general average of all sub-conductors of each line
type. Notice that even though the VX asymmetric
tower has the lowest maximum gradient value, the Cat
face tower presents a general average which is slightly
lower.
The Monopole Tower uses guyed supports with a

single lattice pole and structural parts between phases.
Nevertheless, the spacing between phases has been
minimized for the triangular phase arrangement. To
produce a natural capacity of 1200 MW, an expanded
square bundle, 0.90m/side is used. In this case, the SIL
gain was reached as a combination of a light compaction
(higher Zm) and the use of a semi-expanded bundle
(lower Zs).
5. Comparative analysis of
audible noise (an) and radio
interference (ri)
4 Comparative analyses of the superficial electric fields

The following graphs, the electric field on each subconductors surface, are calculated for a 500 kV line. It is
important to note that when these fields are higher, they
result in more corona activity and, hence, in higher levels
of audible noise (AN) and radio interference (RI) along
the right of way (r.o.w.) and nearby.
The Figure 8 shows the audible noise (AN) profile of all

lines analyzed. It is noticeable that the values of the AN
profile decrease proportionally to the gradients average
of all the sub-conductors, starting with the highest value
of the Cross-Rope, to the smallest value of the Cat Face.
The AN phenomenon generated by the transmission
lines decreases slowly with distance outside of the line
corridor.
Figure 7 is arranged in descending order for the highest

superficial electric field (gradient) calculated with each
design configuration.
Figure 10 Graphic with the electric field profiles
Figure 11 Graphic with the magnetic field profiles
67
500 kV TL Parameters
R1 (/km)
Positive Sequence
Zero Sequence
Ratio
VX asymmetric
VX Symmetric
Cat face
Racket
Cross Rope
Monopole
0.0171
0.0174
0.0175
0.0173
0.0173
0.0176
X1 (/km)
0.269
0.261
0.268
0.267
0.268
0.265
C1 (nF/km)
16.57
17.07
16.56
16.29
16.36
16.63
SIL (MW)
1205
1240
1205
1199
1196
1214
R0 (/km)
0.349
0.349
0.351
0.373
0.346
0.369
X0 (/km)
1.336
1.342
1.329
1.478
1.496
1.361
C0 (nF/km)
9.840
9.764
9.708
7.620
7.251
9.347
X0/X1
4.98
5.14
4.96
5.55
5.59
5.13
C1/C0
1.68
1.75
1.71
2.14
2.26
1.78
Table II: Positive and Zero Sequence parameters of 500kV TLs
The Figure 9 shows the Radio Interference profile (RI)

for the lines analyzed. The phenomenon of RI generated
by the transmission lines decreases more rapidly with
distance than Audible Noise.
r.o.w. The other towers have a higher value on the axis,

with variations in the intermediate area, and significant
reduction to the border of r.o.w.
These calculations show that the triangular phase

arrangements generate higher RI values under the line
but lower levels at the edge of right-of-way. 35 meters
from the axis, the Racket, Monopole and Cat Face
designs have a RI value smaller than the others. The VX
Asymmetric design has smaller RI values than the Racket
in the middle area but levels that are 2 dB/1microV/m
higher on at the edge of right-of-way.
7. Comparative analysis of the

electric parameters
As noted before, the techniques of TL compaction or
TL bundles expansion, leading to an increase of their
SIL, have an impact on the values of their electrical
parameters of the positive and zero sequences (see
Table II). It is observed that the values of reactance (X1)
and capacitance (C1) of the positive sequence of each
configuration are very close, since all designs have been
adjusted to a same value of SIL (1200 MW).
6. Comparative analysis of the

E&M fields near the ground
On the other hand, it is observed that there are major

differences in the zero sequence. Among the reactance
(Xo) there are variations between the minimum (blue)
and maximum (yellow) value on the order of 13%. In
capacitance (Co), the variation reaches 36%. In the
relations Xo/X1 and C1/Co the variations are similar.
This indicates that the performance under unbalanced
conditions and switching transients will have different
responses depending on the maneuver considered and
the adopted tower.
All the field calculations, whose profiles are shown here,

were made at 1.5 meter above ground level. The minimum
phase height was approximately 10 meters to ground, to
allow comparison of line designs. The recommendations
of the ICNIRP limit values or other international bodies
were not considered since these values must be met
at time of executive project and according to the local
regulations.
In the electric field profile, Figure 10, it is verified that
the more compact towers, i.e., the Racket and the CrossRope, present smaller field values in the middle and side
areas of the r.o.w. The Cat Face and the Monopole tower
have the smallest field values in the line axis, getting
higher at 12m from the axis, and having slightly higher
values than the compact towers at 35m from axis. The
towers of plane configuration and with longer distances
between phases have higher value of electric field under
the phases, with reduced slope until the border, similar
to other alternatives.
8. Conclusion
Six TL design concepts of 500 kV TL have been analyzed,
with Natural Power (SIL) of 1200 MW, and similar
positive sequence electrical parameters for system
studies and steady state conditions.
The zero sequence impedance differs by up to
36% depending on the phases and sub-conductors
arrangements.
In the magnetic field profiles (Figure 11) it is verified that

the compact towers have lower values throughout the
TL electrical studies, such as superficial conductors
68
9. References
gradient, environmental interference levels of audible

noise and radio interference, and electromagnetic fields,
showed the advantages and disadvantages of each design.
[1]
In specific applications, where terrain allow the use of

guyed towers or requires self-supporting towers; or in
areas sensitive to audible noise or electric fields at ground
level, the best alternative may be practical.
[2]
Finally, many references on switching transients studies

shows that there is a strong dependence of surge levels
of switching procedures (TL energizing and reclosing,
circuit breakers opening under faults, load rejection,
transient recovery voltage) with the sequence parameters
of the TL nearby. This fact suggests further investigations
of these phenomena in order to have an additional
comparison of the behavior of the concepts of lines
presented in this article.
[3]
[4]
Alexandrov, Georgij N., et alii - The Increase of Effectiveness of

Transmission Lines and Their Corridor Utilization - Cigre Paper
38-104, Paris 1996.
Fernandes, Jos H. et alii Towers for Compact TL of the Second
Circuit of the 500 kV North-Northeast Interconnection at
Eletronorte Electric Studies - VIII SNPTEE 1986 (Paper in
Portuguese).
Regis Jr., Oswaldo; Dart, F. C. et alii Studies and Application of
Expanded Bundle in 500 kV TL - XIV SNPTEE - 1997 (Paper in
Portuguese)
Machado, Vanderlei G. et alii - 500 kV TL of Third North /
South Interconnection Tower Solution with Monomast Guyed
Support and Expanded Bundle - XIX SNPTEE 2007 (Paper in
Portuguese)
69

CSE002

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

CSE002

Hochgeladen von

Copyright:

Verfügbare Formate

Volume N2, June 2015

In this Issue, Best of CIGRE 2014 Session Papers

CIGRE 21, rue dArtois, 75008 Paris - ISSN : 2426-1335

You are holding in your hands the second ever issue

Last but definitely not least, this issue is hosting a

The second group of papers in this issue is devoted

Finally, let me thank you for your interest in CSE

Cigre Science & Engineering

Contents of this issue

Effects of the modularity in PMSM synchronous machine behaviour

Security of communications in voltage control for grids connecting Distributed Energy

Energy Storage Systems Optimization

Wind Induced Motion On Bundled Conductors

Wind Induced Motion On Bundled Conductors

Evaluation of High Surge Impedance Loading (HSIL) solutions

Cigre Science & Engineering

Figure 10: Overview of Simulated Active Power Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 1: Voltage Control use case Function Input/Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 1: Illustration of the storage system principle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Cigre Science & Engineering

Cigre Science & Engineering

Effects of the modularity in PMSM

the generator mounted in a nacelle that sits at the top of a

On the other hand, a high pole number with conventional

Slot per pole per phase

Stator slot height

Maximum ratio L/D

Although required dimension in PMSG are smaller than

Slot/tooth width ratio

Table I. Input data for the analytical iterative procedure.

This paper presents the study of the influence of rotor

Section 2 of this paper present the PMSG geometry of

Magnet pole arc

Rotor external diameter

2. Definition of analysed PMSG

Rotor internal diameter

The PMSG geometry has been developed following an

Rated Power (objective)

Rated voltage (objective)

Table II. Output data from the analytical iterative procedure.

3. Results of the FEM analysis

Cigre Science & Engineering

modelled in FLUX 2D indicating the different parts of the

Figure 1 : Section of th modelled PMSG

To determine the number of rotor modules, it must be taken

Figure 2. Magnetic flux density comparison between no modular rotor (a)

Figure 3 compares the non-modular and the 4 magnets/

Table III . Rotor modularity options

3.1 Magnetic flux density distribution in no load

For modular rotor with five magnets per module, behaviour

Cigre Science & Engineering

magnetic flux density comparison between non-modular

a fractional-slot tooth concentrated winding.

Magnetic flux density in the gap decreases in the area around

3.3 Cogging torque depending on the number of modules

Induced voltage depends on the gap magnetic flux density

Cigre Science & Engineering

the stator slots and it is analysed by means of no load tests.

number of slots and the number of poles of the machine. In

Cogging torque is position dependent and its periodicity

In modular rotor machines, gaps between rotor modules

Cigre Science & Engineering

In these figures, it can be observed that a new component

Table IV. Torque period due to modules of the rotor.