Professional Documents
Culture Documents
CSE007
CSE007
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 CIGRE’s latest publication.
Editorial Committee
Chief Editor Konstantin O. PAPAILIOU
Editors Pouyan POURBEIK
Nikos HATZIARGYRIOU
Rob STEPHEN
You have in your hands the Not only that, In this issue we have included
7th issue of CIGRE Science & for your benefit a paper we believe deserves to
Engineering which traditionally override our internal rule, that only “Best of ”
includes the “Best of ” papers of papers should be included. I am talking about
the Paris 2016 Session. Now let me confess to you, the paper on ‘Using the flow-based method for
that this “exercise” was born two years ago out network investments’ evaluation – a new concept
of necessity! = At the time, there were no paper by Dr. Susana de Graaff, Chair SC C2, and her
submissions -remember this was a brand new co-authors on a subject we have judged to be so
Journal- and we were eager to kick-off. So the innovative and eagerly anticipated by the scientific
idea was born to ask the Study Committee Chairs community, that we did not want to wait until our
to select the three best papers from the papers next issue in June 2017 in order to publish it. I am
approved for the Session and the editorial board sure you will understand and value this decision.
would select the best of the best and publish it in
Before closing let me make a personal statement:
CSE. And so we did!
I believe any technical body, like CIGRE, can
What had started as a convenient solution consider itself privileged, when their top officers
became a “hit” with the authors of the Session are so technically “fit” that they qualify as authors
papers eagerly hoping to “score” and the Journal in some of the “Best of ” papers. I am talking about
readers equally eagerly looking forward to seeing the contributions of our President Dr. Rob Stephen
the results of this unwritten competition. No (‘Calculation accuracy of high-temperature sag
wonder; this selection process has resulted in some for ACSR in existing lines’) and our Chairman of
outstanding papers, like the already stringently the Technical Council Mark Waldron (‘Reduction
selected papers which make it to the Session (firstly of greenhouse gases in GIS pilot project in UK’).
by the respective National Committee, then by the Bravo!
Study Committee Chair and lastly, but not least, by
the Technical Council Chairman) only one paper
Konstantin O. Papailiou
per SC is selected for publication. If I have got my Chief Editor
math right, this comes to an acceptance rate of 1%! konstantin@papailiou.ch
For proof, just take a look for yourself at the
contents of this issue. You will find some remarkable
papers on ‘Evaluation of the thermal performance
of transformers’, ‘Operating experience with and
future challenges for coordinating power plant and
power system protection’, ‘Solid-gas insulation in
HVDC gas-insulated system’ and many others.
page
Editorial
*Juergen.R.Weidner@siemens.com
KEYWORDS
Electrical Lifetime, Hydro- and Turbo-Generators, Silica-Nanoparticles, Stator Winding Insulation System, Treeing
Retardation
Voltage endurance tests at stator winding bars with step away from commercialisation of the new improved
vacuum impregnated epoxy-mica insulation system nano insulation system.
showed that the dielectric strength increases with mica
content up to an optimum [3]. The analysis of many 2. Preparation of nanoparticle
break down channels demonstrated clearly that the loaded epoxy resin
treeing path at the resin to mica interface should be
as long and narrow as possible to get a high electrical The nanocomposite processing technology has a
break down strength in the insulation sleeve [1, 4]. Or tremendous influence on the electrical and mechanical
with other words, to get a long electrical life time of the performance of the insulation system to be developed
insulation system. into an industrial product [5]. Therefore high attention
In addition to these well-understood design measures, was given to find the adequate processing method for the
a significant retardation and lengthening of electrical epoxy nanocomposite. The nanoparticles were prepared
tree propagation through the main wall insulation as alkaline aqueous colloidal silica solution by so-
could be reached by using specially treated inorganic called water-glass process starting from sodium silicate
SiO2 nanoparticles within the resin matrix as shown in chemistry. The Silica-sols were optimized for organic
Figure 2. This paper presents the positive influence of surface treatment and for maximum stability in organic
nanoparticles on the performance of high voltage main Epoxy type systems. Utilizing the correct particle and
insulation for generator stator windings stressed by surface treatment chemistry is very important for long
operation at highly flexible grid demands. It is an actual term stability with regards to changes in viscosity and
status report of the ongoing R&D project, which is still a optimized average particle size.
Figure 2: Delay mechanism of tree propagation throughout the main wall insulation by using small
inorganic SiO2 nanoparticles within the epoxy resin
the tree channel by erosion of epoxy resin. This carbon erosion of polymeric epoxy by PD attack in the channel.
coating could explain the black coloration of the broad Furthermore there are µm-size agglomeration structures
main treeing channel. Beneath this fleecy coating, at the inner wall side, formed after permanent PD activity.
the treeing wall has a smooth surface without any The sequence of PD erosion at epoxy nanocompesites is
macroscopic structures. demonstrated step-by-step in Figure 5 which compares
the SEM pictures (right) with sketches of nanoparticle in
The channel surface at the nanocomposite material is
an epoxy matrix (left).
covered with nanoparticles that are half duged in the
wall surface (see Figure 4 right). The SiO2 particles act More details on investigation and interpretation of
as an inorganic shield like mica does and avoid direct treeing phenomena are given in [7] and [8].
(a) VEI-Factor: 2.0 @ 3-times Un (b) VEI-Factor: 5.1 @ 2.5-times Un (c) VEI-Factor: 14.6 @ 2-times Un
Figure 7: Weibull distribution plot of voltage endurance test results of MICALASTIC system and new nanocomposite system at constant voltage level of (a) 3.0Un,
(b) 2.5Un and (c) 2.0Un
4.2 Voltage endurance tests at original hydro lifetime (voltage endurance) test. Due to the high AC
generator stator winding bars test voltage of 2.5Un = 34.5 kV an accelerated cooling of
the grading system (grey varnish) with fans is needed to
4.2.1 Test specimen
avoid overheating and electrical flash over.
Forty stator winding bars with the original cross The dissipation factor measurement, which had been
section of an indirect cooled hydro generator of rated used as a quality measure for VPI production process,
voltage Un = 13.8 kV and 1.6 m straight bar length were resulted in the same low tip-up values for the bars
manufactured with standard fine mica glass fabric tapes with standard MICALASTIC insulation and with new
and original hot curing VPI process called MICALASTIC nanoparticle loaded insulation system. This clearly
insulation system. For voltage endurance comparison demonstrates the good impregnation of the thicker
tests different sets of bars with pure epoxy resin and with insulation of original 13.8 kV stator bars. All tested
new nanoparticle loaded epoxy resin were produced. bars had a maximum dissipation factor tip-up below
The photo in Figure 10 gives an impression of the high 1.5·10-3/0.2UN. At the new international standard IEC
voltage test cell with a set of these bars connected to 60034-27-3:2015 the tip-up limit is set to a much
high potential of 2.5 times rated voltage for an electrical higher value of 5·10-3/0.2UN.
Figure 11: Weibull probability plot of voltage endurance tests at RT with standard insulation system MICALASTIC and new
nanocomposite insulation system at 3-times and 2.5-times rated voltage
4.2.2 Voltage endurance test field strength of original stator windings, it is mandatory
to qualify the new nanocomposite insulation system
The results of the electrical endurance tests at
according to international standard IEC 60034-18 series
different voltage level are given as Weibull probability
[11].
distribution plots in Figure 11. The Voltage Endurance
Improvement (VEI) Factor verifies the above at tube Therefore the next steps of this development project will
insulation specimens described nanoparticle effect of be:
higher lifetime improvement at lower test voltage. The 1. Transfer the qualification test program also to full-
VEI-Factor increases from VEI = 2.8 at 3Un = 41.4 kV size generator bars with a main insulation up to rated
up to VEI = 4.3 at 2.5Un = 34.5 kV. The 2Un tests with a voltage of Un = 22 kV and a straight bar length of 3
sufficient number of specimens are still ongoing because - 6 m depending on generator type. The qualification
of long lifetime at the lower test voltage. procedure of the new candidate insulation system
4.2.3 Electrical lifetime curves compared to the well proved and established
insulation system has to be performed according
Nevertheless a first electrical lifetime curve ED = f to IEC 60034-18 series [11] dealing with a set of
(tL) of original hydro generator stator winding bars functional evaluation tests for insulation systems
produced in VPI insulation technique with the existing of rotating electrical machines. This qualification
MICALASTIC system and with the new nano-composite program includes the following expensive and time
resin can be drawn in Figure 12. The insulation system consuming tests:
of the bars is designed for a rated line-to-line voltage of - Thermal aging and classification tests at three
Un = 13.8 kV. different temperatures,
To establish the different slope of lifetime curves of - Electrical aging – voltage endurance tests at a
original stator bars additional voltage endurance tests at minimum of three different levels resulting in
four different voltage levels (field strength ED) of 2Un, lifetime curves,
2.5Un, 3Un and 4Un with sample collectives of minimum - Multifactor aging by applying thermal cycle and
7 test bars are needed. At all a production contingent of electrical stress in parallel to the main insulation,
about 70 original hydro generator stator bars has to be - Thermo-mechanical bending endurance in parallel
manufactured and life time tested. with electrical stress as a multifactor aging test.
5. Next steps of development 2. If the new candidate insulation system with
nanocomposite confirms the improved performance,
After successful verification of enhanced electrical then an indirect air cooled generator will be equipped
lifetime with nanocomposites at hydro generator bars with a new designed stator winding using the much
by more than one decade, if extrapolated to operational thinner, better thermal conductive and more flexible
insulation system with SiO2 nanoparticles. It is insulation and in a much higher electrical strength of
expected that the new nanocomposite insulation will the epoxy-mica insulation system for large rotating
increase the overall generator performance (MVA) of machines.
about 10%. (2) The voltage endurance tests carried out on tube-
3. To better qualify the stator winding for new volatile like insulation system models and original hydro
grid demands generated by renewable energy plants a generator stator bars show a great improvement in
basic research project has been started to use special the electrical lifetime of the new nanoparticle based
elastic nanoparticles to make the main insulation high voltage insulation system. The new insulation
system and the winding structure more flexible and system would allow a more efficient generator
to improve dynamic damping characteristics. stator winding design with increased performance in
terms of MVA by about 10%. The nanocomposite
The described long term R&D project dealing with the insulation system will be more robust to peak load
basics in defining applicable nanocomposites usable for operation at highly volatile grid demands with load
high voltage insiulation of stator windings is driven by a ramps up to 100 MW/min in the near future.
well structured technology development process. This step
by step revied R&D project leads to enhanced performance (3) The newly developed nanocomposite insulation
of stator windings in hydro and turbine generators facing system offers a more effective design of stator
the severe stresses of new flexible grid demands. Figure windings. It will result in a smaller thickness of the
13 shows the road map of the project, which started with insulation layer, leading to more space for copper
basic research on different kind of nanoparticles in 2006 material and a better heat transfer. This improves
and is now focused on functional testing of original stator the specific power output and lifetime of new and
winding bars before the last step of prototype production existing (after rewind) hydro and turbine generators.
and fleet introduction could be fulfilled.
7. References
6. Conclusions [1] R. Vogelsang, “Time to breakdown of high voltage winding
insulation with respect to microscopic properties and manufacturing
Based on the promising results of basic material qualities”, PhD dissertation, ETH Zurich, Hartung-Gorre Verlag
Konstanz, 2004.
investigations and further on of insulation system testing
at generator stator bars the following conclusions can be [2] R. Vogelsang, K. Fröhlich, B. Fruth, “Detection of electrical tree
propagation in generator bar insulations by partial discharge
drawn: measurements”, 7th Intern. Conf. on Properties and Applications
of Dielectric Materials, ICPADM June 2003, Nagoya, Japan, pp.
(1) The nano-sized SiO2 particles act like a barrier to 281 – 285.
the treeing process and hold up the treeing channel
[3] A. Wichmann, “Reliability and Testing of High-Voltage Stator
propagation. Obstruction of the tree growth results in Insulation for Large Rotating Machines”, IEEE Winter Meeting
a significant longer lifetime until breakdown of the 1972, New York.
1
University of Stuttgart, Germany
2
GE Grid Solutions, Germany
3
GE Grid Solutions, France
*stefan.tenbohlen@ieh.uni-stuttgart.de
KEYWORDS
Hot Spot Temperature. Oil Flow Rate, Power Transformer, Thermal Modelling
Starting from that design, both an experimental and operating range of a typical desktop workstation, this
two numerical models are created, allowing a detailed contribution specifically investigates the implications of
investigation of the temperature distribution inside the this simplification. For that reason, a two dimensional
modelled disc winding. While measurements, assuming (2D) and a three dimensional (3D) winding model are
a well-controlled experimental setup with negligible investigated alongside the experimental analysis.
measuring errors, offer a dependable and direct insight
into the investigated physical phenomenon, they are 2. Development of an
also afflicted with several substantial disadvantages.
Considering a strongly limited model flexibility
experimental model for a disc
alongside substantial time requirements and high costs, winding
the development of alternative investigation concepts
The design of disc type transformer windings shows a
appears increasingly advisable.
strong symmetry in circumferential direction. To display
Consequently, two numerical winding models based on this design characteristic more clearly, Figure 1a)
computational fluid dynamics (CFD) were created. Due to offers a three dimensional view on a disc type winding
the considerable computational requirements connected with four conductor turns per disc. For reasons of an
to CFD analysis, most CFD-based investigations of improved visibility concerning the cooling system inside
transformer windings apply a reduction of the winding the winding, the outer cardboard cylinder is displayed in
geometry to a two-dimensional (2D) domain. While a partly transparent color. The given enlargement shows
this reduction appears to be a suitable strategy to the horizontal cooling channels created by the opening
keep the connected computational efforts within the between two consecutive spacers and discs. The vertical
Figure 2: Design characteristics of the modelled winding geometry and the winding model [3]
cooling channels are bounded by the cardboard cylinders to identical amount of conductor surface available for
and by the sticks aligning the spacers. Consequently, heat exchange (marked green) and conductor surface
the entire winding geometry can be subdivided into covered by spacers (marked red). Therefore, the thermal
symmetrical sections, as separately displayed in and hydraulic impact of the design modifications is
Figure 1b). To take advantage of this geometrical considered negligible.
symmetry for the heat and mass transfer analysis, the
Finally, the paper wrapping typically applied around
following premises have to be met.
conductors in transformer windings for purposes of an
• Homogeneous inlet conditions below the winding electrical insulation is substituted by solid blocks of
• Homogeneous exit conditions above the winding plastic (Polyvinylidenefluoride) located only in between
• Evenly distributed electrical losses within each the conductors of a disc. Prior experimental investigations
conductor turn showed a great sensitivity of the evolving temperature
Although not every transformer design fulfills these levels on the exact thickness of the applied paper
premises, they are assumed to be valid within this wrapping. Unfortunately, this property, especially with
contribution. Consequently, all windings models regard to the comparably short experimental conductor
investigated below represent one symmetrical section of models, is difficult to keep within small tolerances. To
a disc winding. eliminate respective detrimental influences but still
allow a realistic thermal decoupling of the conductor
As carried out in the introduction, a winding design models in one disc, the paper is substituted, as illustrated
with characteristics being representative for an OD- in detail in Figure 2c).
cooled disc winding is chosen as a common basis for all
investigated winding models. A top view on a symmetric The design of the single conductor models, arranged
section of this chosen winding design, including all into discs in the further assembly process, is displayed
relevant geometric details to assess this design, is in Figure 3. To provide the heating power normally
shown in Figure 2a). Since the exact reproduction of dissipated in the winding turns of a power transformer
this geometry would pose enormous challenges for during operation, heating cartridges are injected into
the connected manufacturing processes, the respective through bore-holes. An accurate measurement of the
design of the experimental model was modified. For individual conductor temperatures is accomplished with
that purpose, the curvature of the conductors in temperature sensors located inside a second drilling, as
circumferential direction is neglected, as displayed in indicated in Figure 3a).
Figure 2b). The cooling system of OD-cooled disc windings is
Furthermore, the symmetry of the horizontal cooling usually subdivided into so-called passes in axial direction.
channels and spacers is forfeit. The reason for this By periodical placement of washers inside the vertical
modification results from the material choice for the channels alternating between the inner and outer winding
smaller and rectangular spacer shown in Figure 2b). diameter, two different types of passes result. Therefore,
To allow a future optical investigation of the oil flow the experimental winding model must consist of at least
distribution with the experimental winding model, two consecutive passes. In addition, representative
these components were made of glass. Due to the far entering and exiting conditions are required before and
more complex manufacturing processes connected to after the investigated passes. This ensures thermal and
that material, a rectangular shape was chosen as a cost hydraulic conditions that are representative for every
effective compromise. Nevertheless, both geometries pass throughout the complete cooling system of the
given in Figure 2a) and Figure 2b) share the identical modelled winding. Integrating these considerations,
opening widths of the horizontal channels and a close Figure 4a) shows the chosen winding model layout in a
schematic view. To its right, Figure 4b) provides a clear heater, the initially provided oil flow rate from a gear
reference for the pass, disc and conductor numbering the pump is limited to a defined value by a controlled valve
presented post analysis relies upon. Finally, Figure 4c) processing data from a flowmeter. To minimize the heat
contains a photo of the experimental winding model leakage from the winding model to its surroundings, the
during operation. Although this picture gives no detailed entire model is submerged in an open experimental tank
insight into the chosen mechanical design of the winding filled with oil of the same temperature as provided by
model, comprising its sealing and insulation concept, it the flow heater. Finally, the individual power supplied
already offers a first impression of the required technical to each heating cartridge is set by software controlled
efforts. power electronics. Due to the chosen laboratory setup,
Next to a representative winding design, also the a total measuring tolerance concerning the temperatures
operating conditions of the modelled winding are of approximately 0.5 K is achieved. Consequently,
chosen accordingly with respect to the winding type and temperature gradients can be determined with an
dimensions. To control the relevant boundary conditions accuracy of around 1 K. The occurring deviations
precisely, a laboratory setup schematically illustrated in concerning the mass flow measurement and heating
Figure 5 was applied. It allows an accurate control over power control are below 1 %.
the mass flow rate , the oil temperature at the model
inlet and the heating power of each conductor Plosses, i.
3. Development of CFD models
While the oil inlet temperature is set via a controlled flow for a disc winding
Figure 5: Piping and instrumentation diagram of the laboratory setup designed to create strictly controlled bound-
ary conditions inside the winding model [3]
Figure 7: Comparison of the oil flow velocities within the 2D and 3D CFD winding model at
*reinhard.goehler@siemens.com
KEYWORDS
MO arrester – metal-oxide resistor (MOR) – HVDC – valve protection – thermal behavior – equivalent continuous
operating voltage (ECOV) – energy handling – creepage distance – insulation withstand
absorption capability of the arresters (their repetitive by simply omitting the housing. To further improve an
charge transfer rating, respectively, acc. to the arrester arrester’s thermal behavior, enforced heat transfer can
standard IEC 60099-9 [4]). be considered by using e.g. compressed air of controlled
temperature or water cooling as it is used for thyristor
After all the design process is an optimization between the
cooling. Both will be available in converter halls, but it
valve arrester and the thyristor valve design parameters
is very complicated including these measures into the
(e.g. number of series connected thyristors and layout of
arrester. A specific design of passive cooling elements is
the snubber circuit).
much easier to handle in the overall arrester design. In
It is obvious that in general valve arresters with increased addition to the cooling improvement specific creepage
energy absorption capability render possible designs distance requirements have to be fulfilled.
with lower number of series connected thyristor levels.
Therefore, an arrester design without housing and 2.1. Design of insulation
additional design features for better heat transfer has According to IEC 60071-5 [3], a creepage distance of
been introduced, designated as “High Cooling Arrester” 14 mm/kV is required for indoor insulation. This means
(HCA). The pilot application has been the HVDC system for the MO resistors used for the HCA application that
Storebaelt where the arresters are in operation since 2009. their intrinsic creepage distance (i.e. the direct path on the
Ever since the HCA has been installed in a growing number coating between the metallic end faces) must be doubled.
of valves in HVDC systems without any problems. This is achieved by sheds made of silicone rubber, which
Under certain circumstances the protective level of a valve are slipped over the MO resistors as can be seen in Fig. 4.
arrester can be reduced by more than 10 % compared to an
arrester with housing, and the number of series connected
thyristor levels can be reduced accordingly with the
corresponding effect on costs. For the back-to-back
system Hudson the protective level of the valve arrester
could be reduced by 10 %. Since only discrete numbers
of series connected thyristor levels can be realized within
the thyristor modules this resulted in 13 % reduction of
the number of thyristor levels. A multiple valve unit in the
Hudson system comprises four valves belonging to one
AC phase of the 12-pulse group. In this arrangement it Figure 4: MO resistor with slipped-over silicone rubber shed and cooling fins
was possible to reduce the number of thyristor modules
from 6 to 4 resulting in a height reduction of the multiple 2.2. Design of active cooling element
valve unit from 9 m to 7.5 m (Fig. 2 and Fig. 3). To find the best geometry of cooling elements, thermal
simulations have been carried out for complete arrester
2. High cooling arrester design assemblies consisting of MO resistors, polymer sheds
Valve arresters without housings have been in service for and metallic cooling parts. The simulations were carried
many years. Their better thermal properties are obtained out with the FEM software ANSYS. In general, heat
transfer is constituted by heat convection, radiation 20 % higher compared to that of each column of HCA1-3.
and conduction. In order to determine the simulation This increase was chosen to achieve the same equivalent
parameters, preliminary simulations and tests on different continuous operating voltage (ECOV), determined
models were carried out. With the final parameter set according to IEC 60099-9 [4], for both arrester designs.
of convection coefficient α, emissivity e, specific heat The tests were performed under applied alternating
capacitance and thermal conductivity of the three materials voltage in a closed test chamber with a volume of 200
– varistor ceramic, silicone rubber and aluminum – the m³. Its internal temperature can be controlled by heaters
geometry of the cooling elements was optimized. The and air condition up to a maximum of 60 °C. The test
simulations were started with simple aluminum parts transformer is located in the chamber. As the test voltage
as can be seen in Fig. 5 a). Different steps enlarging the is limited by the transformer rating, the test samples were
metallic cooling area including horizontal fins lead to only chosen to have an ECOV of 94 kV. The whole test setup
minor improvements. The final step, introducing cooling is shown in Fig. 6.
elements with vertical fins as shown in Fig. 5 b) results in a
cooling behavior which is about 45 % better than the initial
one. Electrical behavior, such as dielectric strength and
absence of partial discharges (electrical field simulation,
partial discharge and withstand tests, see Section 4) and
the reduced overall length of the arrester were taken into
consideration. For evaluation the temperature increase
within the MO resistors was taken at constant power loss
and heat generation in the MO resistors.
3. Thermal investigations
To investigate the thermal behavior of the HCA and to
compare it with that of a conventional housed arrester
(HA) four different arresters are considered:
1. Housed arrester with one MO resistor column (HA1)
2. High cooling arrester with one MO resistor column
(HCA1)
3. High cooling arrester with two MO resistor columns
(HCA2)
4. High cooling arrester with three MO resistor columns
(HCA3)
Figure 6: High Cooling Arrester (HCA3) installed in the test chamber for
The number of MO resistors (MOR) used for HA1 is thermal testing at ambient temperature of 20 °C and 60 °C
Type test
A complete type test according to standard IEC 60099-
9 [4] was successfully carried out using the developed
parameters. The steady state temperature during the tests
performed on real size, completely assembled single-
and multi-column arresters was approximately 80 °C.
Therefore, the preheating temperature during the thermal
recovery test was chosen as 80 °C instead of 60 °C.
6. Bibliography
Figure 12: U-I-characteristic of the MO resistor column used for insulation [1] K. Steinfeld, R. Göhler, D. Pepper, „High Voltage Surge Arresters
withstand test in comparison with the standard protection characteristic for Protection of Series Compensation and HVDC Converter
Stations”, The 4th International Conference on Power Transmission
Summary and conclusion and Distribution Technology, October 11-14, 2003, Changsha,
China
A special design of a valve arrester for protection of [2] CIGRÉ Working Group A 3.17, “MO Surge Arresters - Stresses and
the valves in a line commutated converter (LCC) has Test Procedures”, CIGRÉ Tech. Brochure 544, 2013, pp. 15–23
been developed, designated as “High Cooling Arrester” [3] IEC 60071-5, Ed. 1.0, 2014-10: “Insulation co-ordination – Part
(HCA). The arrester has an extremely low protective 5: Procedures for high-voltage direct current (HVDC) converter
level, which helps saving overall valve costs, but for the stations”
price of high power losses and thus heating under the [4] IEC 60099-9, Ed. 1.0, 2014-06: “Surge arresters – Part 9: Metal-
typical operating conditions. To manage this its cooling oxide surge arresters without gaps for HVDC converter stations”
behavior has been improved not only by completely [5] IEC 60099-4, Ed. 3.0, 2014-06: “Surge arresters – Part 4: Metal-
omitting the housing but additionally by introduction of Oxide surge arresters without gaps for a.c. systems”
*yuji.yt.tateno@furukawaelectric.com
KEYWORDS
Renewable energy - Floating wind turbine - Dynamic cable -Riser cable - Submarine joint
3.2 Cable Behavior Analysis In dynamic analysis, the riser cable configuration, curvature,
maximum load are checked while the floating body in a
The oceanographic condition such as height and frequency
neutral position is affected by irregular waves corresponding
of significant wave and a floating body’s dynamic response
are needed at first for external conditions. to wave-fluctuating amplitude. It is necessary to verify the
following aspects in the riser cable configuration: “The
In addition to the basic physical properties of cable catenary lowest point is above the ground.” “DBMs are
diameter and weight, and the mechanical parameters of below the surface of sea.” “The ground touching position is
cable’s longitudinal rigidity (EA), bending rigidity (EI) and within the protected section.” It is also important to confirm
torsional rigidity (GJ) are required for the behavior analysis. that the tension and the curvature at the top of riser cable
The bending rigidity is the most important to check the are within allowable values. Fig. 7 shows simulation results
cable curvature. As the mechanical parameters the actual of dynamic variation for tension and curvature at the top
measured values have been used in those analyses. The of riser cable. One of difficult aspects was that the floating
analytic model of Lazy-Wave is shown in Fig. 5 body’s moving distances were larger than the water depth
In static analysis, the riser cable configuration, curvature, in some cases.
maximum load (tension, shearing force, bending moment)
while the floating body has a maximum displacement
with a maximum current are checked. Fig 6 shows some
simulation results of riser cable configuration in case
that the floating body moves towards three directions
from a neutral position.
Nominal
66 kV 22 kV
voltage
Conductor 3 x 100 mm2 3 x 150 mm2
Insulation
11 mm 6 mm
thickness
Double layers of galvanized steel
Armor
wires (6.0 mm)
Outer diameter 175 mm 147 mm
Item Requirements
Nominal voltage 66 kV 22 kV
Connection It enables to connect a submarine cable and a riser cable.
- AC: 130kV x 3 hours (at R.T.) - AC: 57kV x 3 hours (at R.T.)
Withstand Voltage - Lightning Impulse: - Lightning Impulse:
±485kV x 3 times (at R.T.) -230kV x 3 times (at R.T.)
130m of maximum depth
Bearing Water Pressure 130m of maximum depth Water permeability
≤ 1×10-7 [g∙(cm/cm2)∙day∙mmHg]
Bearing Tensile Strength 93.1kN approx. of immersing tension
Assembling It can be assembled on the vessel.
7. Conclusion
The cable installation and connection to the world’s
largest floating 7 MW offshore wind turbine in the 2nd
stage were successfully completed, and it is already
under operation.
There is, so far, not enough experience in riser cable
transmission system which is required for the floating
power generation system. With the facts found
throughout the Fukushima FORWARD Project, the best
transmission system design and the maintenance method
will be established.
8. Acknowledgments
This research has been carried out as a part of Fukushima
floating offshore wind farm demonstration project
funded by the Ministry of Economy, Trade and Industry.
The authors wish to express their deepest gratitude to the
Fig. 13 Installation procedure of 22 kV riser cables and submarine cable concerned parties for their assistance during this study.
Abstract temperature. Since, above the KPT, the rate of sag increase
with temperature is reduced, the determination of KPT is
According to a CIGRE questionnaire completed in a significant factor in evaluating line uprating methods
2003, over 80% of existing lines were built with steel using the existing conductors.
reinforced aluminium conductors (ACSR). Therefore,
when attempting to increase the thermal rating of existing If the aluminium layers of ACSR undergo plastic
elongation due to metallurgical creep or high tension load
High Voltage (50 kV to 345 kV) transmission lines, low-
events, then the knee-point temperature of the conductor
cost line uprating methods often involve allowing the
is reduced, typically decreasing over the life of the line.
existing ACSR conductors to operate at an increased
In calculating the conductor sag at temperatures above
Maximum Allowable Conductor Temperature (“MACT”)
the KPT, the possibility of compression forces, residual
temperature. Of course, minimum electrical clearances to
manufacturing stress, and greater than anticipated plastic
ground, buildings and other lines must be maintained and
elongation of the aluminium layers must be considered.
the conductor system unharmed by operating at higher
temperatures. If these criteria are not met, the line must be 1. Introduction & Background
reconductored or rebuilt.
In designing a new overhead transmission line, one of
For ACSR, all sag-tension calculation methods assume the most basic and necessary calculations is that of sag-
that the unstressed length of aluminium and steel layers tension. CIGRE Technical Brochure 324 [1] explains
is the same before and after installation. After initially the calculation of sag and tension for bare overhead
sagging the conductor, high line currents cause the conductors under various ice/wind conditions and at
conductor temperature and sag to increase. CIGRE the high conductor temperatures produced by the line’s
Technical Brochure 324 suggests two linear and one highest power flow. Based on these calculations, span
non-linear conductor model. The Experimental Plastic lengths and structure heights are chosen to meet the
Elongation (EPE) model considers the aluminium layers structure loads and electrical clearance requirements over
and steel core separately. The paper explains and quantifies the life of the line.
the various factors that influence the calculation of ACSR
sag at high temperature. In evaluating existing lines, power flow constraints
on the AC Transmission System are often the result of
The thermal elongation rate of the aluminium layer(s) is inadequate thermal rating on older existing High Voltage
twice that of the steel core, and, while the total tension (50 kV to 345 kV) transmission lines, built with one
decreases with increasing temperature, the tension in the ACSR conductor per phase. Increasing the thermal rating
aluminium layers decreases faster than the tension in the of these existing lines can often be accomplished by
steel core. For strong ACSR conductors in short spans increasing the maximum design temperature of the lines
operated above 75oC, the aluminium layer tension may if this can be done while continuing to maintain minimum
reach zero at a “knee-point temperature” (KPT) which electrical clearances to ground, buildings and other lines.
is less than the line’s actual or proposed templating As described in [2], if the sag of these ACSR conductors
* Da.douglass@ieee.org
KEYWORDS
High temperature sag, knee-point temperature (KPT), aluminium layer axial compression, ACSR, LiDAR, Linear
Elastic, Simplified Plastic Elongation, Experimental Plastic Elongation.
Table 1: Sag Errors at High Temperature for three Different ACSR Strandings
and that (2) the length of the aluminium layers and the With ACSR, the core usually undergoes only a small
core must remain equal. For an aluminium conductor amount of plastic elongation during high load events and
with a steel core (i.e. ACSR), suspended in a catenary, the the impact of plastic elongation in the aluminium layers is
aluminium strand layers of ACSR elongate at twice the different. Using the EPE stress-strain model, the amount
rate of the steel core. of plastic elongation in the aluminium layers can be
calculated as a function of both high tension load events
If the conductor is modelled with the EPE method, the
and creep elongation over time. All three stress-strain
tension of the steel core and the surrounding aluminium
models allow one to calculate the KPT of ACSR, but LE
layers are modelled separately. Therefore, as the ACSR
assumes that it remains constant over time and loads,
conductor temperature increases, the total conductor
SPE assumes that KPT decreases a set amount over time,
length and sag increase and the total conductor tension
and EPE can be used to calculate the KPT of ACSR for
decreases. However, for the lengths of the steel core
user-specified loading events and, given creep elongation
and the surrounding aluminium layers to remain equal,
models, the variation of KPT over the life of the line.
the greater thermal elongation of the aluminium must
be offset by reduced elastic elongation and the lesser Other factors can also influence the KPT. The initial
thermal elongation of the steel core must be increased KPT of ACSR (or other reinforced conductors) can be
by increased elastic elongation. In other words, as the reduced by pre-stressing the conductors prior to sagging
conductor temperature increases, the percentage of total and clipping. Rawlins [7] suggests in some detail that the
tension in the aluminium must decrease. KPT can be higher than expected due to residual stranding
stresses in the aluminium wires of ACSR made in modern
At a temperature referred to as the “knee-point
rigid-frame stranding machines, which effectively makes
temperature” (KPT), the tension in the aluminium layers
the zero-stress aluminium wire layers shorter than the
of the ACSR conductor has decreased to zero (see Figure
zero stress length of the core. Finally, long pulls and
1) and all of the conductor tension is in the steel core.
high tensions during tension stringing can produce non-
Above the KPT, the aluminium wire layers go into
negligible plastic elongation of the aluminium prior to
compression. If the elastic modulus of the aluminium
sagging. These phenomena are best studied using the EPE
layers in compression is equal to that in tension (think
model.
of the behaviour of Aluminium-clad steel wire), then the
CTE remains equal to that at temperatures below the KPT. Above the KPT, the continued thermal expansion of
If the elastic modulus of the aluminium wire layers is aluminium yields increasing aluminium compression but,
zero, then the conductor’s composite thermal elongation if the aluminium compression modulus is low, the rate of
rate decreases to that of the steel core alone, depending sag increase with temperature is sharply reduced to near
upon the compression modulus of the aluminium layers. that of the steel core alone as shown in Figure 2.
In any stranded aluminium conductor, the aluminium The advantage of using the EPE conductor model is that
wires undergo both elastic and plastic elongation. With it yields lower calculated high temperature sags (about a
an all-aluminium conductor, plastic elongation due to meter less at 150oC as shown in Figure 2). On the other
high load events and metallurgical creep, cause both the hand, because it is not as conservative as the LE/SPE
everyday sag and the high temperature sag to increase conductor model, there must be increased concern with
over the life of an overhead line. those factors that affect the modelling.
There are three primary concerns with regard to sags 6. Compression of helical
calculated by the EPE conductor model. They involve the
conductor manufacturing details (lay lengths and bobbin aluminium strand layers
tension), the conductor installation procedures, and the For single-layer and multi-aluminium layer ACSR
occurrence of severe mechanical ice & wind load events conductors, at conductor temperatures above the KPT, the
prior to the field sag measurements of the existing line. aluminium wires continue to expand faster than the steel
The KPT depends on the ACSR stranding (Figure 3) and the core wires putting the aluminium layers into compression
effective span length (Figure 4) as summarized in Table 2. and increasing the elastic elongation of the steel core.
Given the physics of the ACSR conductor, aluminium
Table 2: “Final” knee-point temperature as a function of ACSR
steel content and ruling span length. compression must occur above the KPT.
Figure 3: Comparison of sag vs temperature for 45/7, 54/7, and 30/19 ACSR Figure 4: Sag vs conductor temperature for 180, 300, and 425 m spans.
with a 300 m ruling span.
Where:
D = Overall diameter of strand layer
Where: E is the axial elastic modulus of aluminium d = Strand diameter in layer
(slightly less than 60 GPa). R = Radial expansion of the strand layer
n is the number of strands in the layer. Considering the number of strands in the layer, n, as a
α is the lay angle of the layer. function of strand and layer diameters:
Eff is a weak function of the lay angle and is typically
slightly less than 4 GPa. Therefore, the axial modulus
of the aluminium layers in compression is less than 10%
of the axial modulus in tension unless the layers are Then:
constrained radially.
Figure 8: Comparison of measured and calculated sag vs temperature for 26/7 ACSR
calculations can be important in establishing the thermal The experiments at both locations indicate that high
capacity of existing clearance-limited overhead lines and temperature sag calculations using the EPE conductor
in determining the best method for increasing the thermal model should use an assumption of 15 to 20 MPa for
capacity of such lines. residual aluminium stress to be conservative on the basis
of this limited experimental data.
Two careful laboratory experiments are discussed in the
following paragraphs. Figure 8, shown in Reference [7], 8. High Temperature Sag in new
is based on tests performed as part of a study of high
temperature thermal elongation with an indoor, 90 meter and existing lines
test span. The conductor is 26/7 Hawk (240 mm2) and it it may be observed that sag at high temperature is seldom
was pre-stressed to simulate the aluminium layer plastic measured in either new or old lines. In both cases, high
elongation that would occur as a result of creep and high temperature sag is calculated based upon sag measurements
load events in an installed line. made at everyday line currents and weather conditions.
This experimental data was analysed by Rawlins [7] and For new lines, field measurements are made after tension
presented in an IEEE paper which elicited a good deal stringing as part of line construction. For existing lines,
of interesting discussion. In essence, Rawlins proposed measurements are made with the line carrying normal
current (or none), by LiDAR or conventional survey
that the experimental data could best be explained by
methods.
“residual stress” in the aluminium layers produced as a
result of stranding in a rigid-frame stranding machine. With new and existing lines, given non-homogeneous
He supplemented this data by measurements of residual conductors (both conventional ACSR and HTLS), the
stress in various ACSR conductors prior to installation tension distribution between core and aluminium layers is
and concluded that the KPT should be adjusted upward estimated at the time of measurement.
by assuming a residual stress of 15 MPa. The EPE analysis presented in this paper is also applicable
A second experimental study was performed in an outdoor to the analysis of high-temperature, low-sag conductors
test line (2 spans) at the Oakridge National Laboratory in having either zirconium or annealed aluminium layers
Oakridge, Tennessee. The analysis was reported by Seppa surrounding a core which has high strength and low
[8] and his plot of the test data is reproduced in Figure 9. thermal elongation.
The test facility consisted of two 180 m spans, carefully
terminated to allow accurate sag measurement and with 9. Selection of High-Temperature
many thermocouples embedded in the conductors. Sag model for new lines
The laboratory experiments demonstrate that the measured when designing new overhead transmission lines,
sags are less than those calculated with the LE or EPE, the height and the placement of support structures is
constant CTE model and greater than that calculated dependent on wind and ice loads (primarily on the
with the EPE/Graphical method where residual stress is conductors) and on the sag of the phase conductor at the
ignored. templating temperature. If the reduction in slope of sag
When calculating the sag of ACSR at conductor 1. The maximum operating high temperature of the line
temperatures above 75°C, the thermal elongation rate is determined by the clearances produced when the
of the conductor should model the mechanical and conductor is at the templating temperature.
thermal behaviour of the aluminium layers and the steel 2. The spans are short (<400m).
core separately. If the conventional ACSR temperature- 3. The conductor has a steel core area greater than 10%
independent CTE formula is used, the sag at maximum that of the aluminium layers.
line temperature will be overestimated. In designing new 4. Lay ratios are near recommended values.
lines, allowing for the modest amount of excess high When the sag of ACSR at high temperature is essential
temperature sag which results from the linear CTE method to the selection of line uprating method, the KPT should
is of minor importance. However, in evaluating options be determined by use of the EPE mechanical conductor
for thermal uprating of existing lines, the difference in model (including the impact of metallurgical creep of
the linear and non-linear CTE methods can have a large aluminium and plastic elongation for high tension load
impact on the uprating method chosen. events) and a residual stress of 15 to 20 MPa should be
If the conductors are intended to operate at temperatures specified for the aluminium layers.
above 75°C, it is recommended that a model be used that
takes into account the knee-point of the conductor. Failure
13. Bibliography
to do this may result in a suboptimal conductor being [1] CIGRE Task Force B2.12.3, “Sag-tension Calculation Methods for
Overhead Lines”, Technical Brochure 324, June, 2007.
chosen for the line.
[2] CIGRE Task Force B2.12, “Conductors for Uprating of Overhead
The temperature gradient also needs to be considered as Lines”, Technical Brochure 244, April, 2004,
the average temperature determines the conductor sag and [3] IEEE Subcommittee 15.11, “Limitations of the Ruling Span Method
not the surface temperature [10]. for Overhead Line Conductors at High Operating Temperatures”.
Report of IEEE WG on Thermal Aspects of Conductors, IEEE WPM
As the conductor temperature increases above ambient, 1998, Tampa, FL, Feb. 3, 1998
the aluminium layer thermal expansion rate is twice that [4] CIGRE WG B2.12, “Alternating Current (AC) Resistance of
of the steel core and at the “knee-point temperature” Helically Stranded Conductors”, Technical Brochure 345, April,
(KPT), the tension in the aluminium layers is zero and 2008.
all the tension is in the steel core. Beyond the KPT, the [5] IEC 1597, “Overhead electrical conductors – Calculation methods
CTE of the conductor is reduced and the conductor sag for stranded bare conductors”, First Edition, 1995.
which corresponds to the line’s maximum temperature is [6] Varney T., Aluminium Company of America, “Graphic Method
for Sag Tension Calculations for A1/S1A (ACSR) and Other
reduced. Conductors”, Pittsburg, 1927
The issue of KPT and its impact on the line sag at high [7] Rawlins, C.B., “Some Effects of Mill Practice on the Stress Strain
temperature depends on a number of line design factors. Behavior of ACSR”, IEEE Transactions on Power Systems, Vol. 14,
No. 2, April, 1999.
The issue is of minimal concern when:
[8] T. Seppa “Fried Wire?” (Public Utilities Fortnightly, December 2003,
1. The maximum operating high temperature of the line is pages 39-41)
determined by loss of strength in the aluminium layers [9] CIGRE Working Group B2.36, “Guide for Application of Direct Real-
rather than sag. Time Monitoring Systems”, Technical Brochure 498, June, 2012.
2. The spans are long (>400 m). [10] Clairmont, B, Douglass, D., “Radial and Longitudinal Temperature
3. The conductor has a steel core area less than 7% that of Gradients in Bare Stranded Conductors with High Current
the aluminium layers. Densities”, CIGRE Paper B2-108, Paris, 2012.
4. Compression modulus of the aluminium layers is
negligible.
*kobayashi.takayuki@tepco.co.jp
KEYWORDS
Asset management, Aging transmission facilities, Health Index, Economical life time, Investment strategy
In the asset management, the data of maintenance/ d. Setting of life expectancy and determination of
inspection and evaluation of current situation which refurbishment and repair
are prepared by Service Provider are utilized by Asset Except for transfomers, the lifetime evaluation method
Manager and Asset Owner in their policy decision. Also, of facilities has not been developed. The lifetime
degree of communication between Service Provider and expectancy of facilities can be empirically estimated as
Asset Owner has to be reinforced. shown in Table 1 [1], but there is a gap between 10 to
40 years against the real value. In addition, quantitative
The challenges in the asset management are identified as
determination criteria tor refurbishment and repair has
follows;
not been developed.
a. Arrangement of Health Index showing condition
Table 1 Lifetime expectancy of each facility [1]
evaluation of each facility
Recently, examples of actual applications examples
have been reported from some countries. Evaluation
items are being identified gradually, however, the way
of weighting them are not developed. Also the Health
Index items of transfomers are practical, whereas those
of circuit breakers are not, due to the lack of practical
external diagnostic technologies. 3. Challenges of asset
b. Collection of fault data for failure rate prediction management in japan
Although the fault records is important, those data can As discussed in the last chapter, there are still challenges
not be shared among users because each of them use in asset management such as development of the Health
original data collecting format. Index system, standardization of data collecting format,
As the failure rate of transmission facilities is very and formulation of a prediction method for failure rate.
low, a large amount of data samples are required to
accurately predict the failure rate in the future. And it is Also, to perform the asset management for substation
necessary to have a standardized format to collect and system as a whole, it is necessary to consider challenges
share failure records. such as the impact on the power system, capability of
c. Prediction of failure rate construction, and maintaining of technical ability.
It is necessary to know the timing and the speed of In Japan, electric power companies, manufacturers,
incline of the failure rate bathtub curve for an accurate universities and research institutes are cooperating to
prediction. However, the prediction method of the failure respond to the challenges mentioned above. We will
rate based on limited data samples is not developed. report the activities in this paper.
failure rate of an operating equipment at the referenced By multiplying repair cost to this result, it is possible to
year, and is approximated by the Weibull distribution as estimate repair cost by CM and evaluate the life cycle
per the following; cost, and in turn leading to strategic decision making.
dR(t)
λ(t) = ¯ ____ /R(t) λ(t): Hazard function 3.4. Asset managemant as a substation system
dt
R(t): Reliability function* When it comes to the asset management as a substation
R (t) = 1 – F (t) F(t): Unreliability function system, not only the current conditions and priorities of
each facility, but the typical challenges in substations
F (t) = 1 –exp{–(t/ts)m} F(t): Cumulative failure must be factored in as we. These challenges include
rate by Weibull function power system factors caused by plan or change in power
m: Shape parameter demand, coordination of the number of field workers and
ts: Scale parameter maintaining technological competance, and the increase
Current facilities have been operating for approximately in obsolete type equipment.
30-40 years and by applying the hazard function to the In Japan, for transformers and early type gas insulated
aging distribution, the future failure rate can be obtained switchgears (GIS), guidelines for life extension, partial
as a bathtub curve. In the case of fault example in Figure repair, and refurbishment of substation system are
2 (c), a rapid increase in faults after about 40 years of determined upon assessment of the substation’s impact to
operation is predicted. the power system.
Figure 4 shows the guideline for the early type GIS adopting this idea to a GIS installed in a Japanese electric
[3]. First step is assessment of equipment condition. company [4]. For each manufacturer’s GIS type, risk
In the second step, life extension or refurbishment is items such as maintainability of aging facilities, presence
decided, considering the possibility of life extension. of implementation of countermeasures to same type
In the next step, it is assessed decided once more from equipment, failure rate, and production of discontinued
the viewpoint of economic evaluation. Refurbishment parts, are evaluated by scoring points in Table 5. In
priority is decided based on evaluation items in Table 4. addition to leveling the number of planned replacements,
In the refurbishment planning step, site condition, system by utilizing the characteristics of each manufacturer’s
reliability, standardization of investment, and capability facilities, an appropriate maintenance is made possible
of construction are considered. based on the facilities’ soundness. Furthermore,
cooperation of facilities countermeasures and regular
4. Examples of asset management maintenance can be realized.
Using the approach reported in Chapter 3, the priority of Moreover, in the future, combining the asset management
transmission facilities can be decided by not only aging methods to the capital management system will link
but also by indexing the condition of facilities, impact of Service Provider and Asset Owner/Asset Manager more
faults and economic loss. Figure 5 shows an example of strongly.
Figure 5 GIS refurbishment schedule based on the risk management in Japanese electric company
1. Introduction
The power supply for both Amazonas and Amapá States,
located in Amazon region - Brazil, was done through
isolated systems under concession of Eletronorte, as
well as distribution and generation entities before 2012.
Several studies have been conducted in recent decades
to incorporate these isolated systems into the Brazilian
Interconnected Power System (BIPS). In 2003, a study
Figure 1 - Power system map of TMMI project highlighting
conducted by the then Planning Authority concluded the different auction’s lots
the attractiveness of the incorporation of these isolated
systems of the Amazon region to BIPS. Techno- A better visualization of TMMI can be shown in Figure
economic feasibility studies concluded later that the best 2. TMMI passes along three federation states, namely:
alternative for integrating these isolated systems was the Pará (PA), Amapá (AP) and Amazonas (AM). The supply
*ricardo.tenorio@ons.org.br
KEYWORDS
FSC - Hunting – Oscillations – Subsynchronous – SVC
of 18 FSCs by a single manufacturer is considered one iv. Silves SVC - 2 TCRs rated at 147.6 Mvar each; 2
of the world's largest to date. The support of 4 SVCs TSCs rated at 129.4 Mvar each; 2 single tuned filters
adds some more control and flexibility for operation but (5th harmonic) at 36.8 Mvar each – 60 Hz and 20 kV
it comes along with some complexity to run and is prone (secondary voltage). Output -200 to + 300 Mvar on 500
to some control instability due to SSO to be discussed in kV, 15% coupling transformer reactance (300 MVA).
this article.
Figure 3 shows the simplified single line diagrams of
Jurupari and Silves SVCs. Note that Jurupari, Oriximiná
and Macapá SVCs, which belong to lots A and B, were
manufactured by one single manufacturer, identified
here as manufacturer I. Silves SVC, which belongs to lot
C, was manufactured by another manufacturer, named
here manufacturer II. Note that Jurupari and Oriximiná
SVCs are of similar type with two TCRs/TSCs, while
Macapá SVC has only one TCR/TSC. All 4 SVCs have
in addition to the main controller (PI controller) some
control features, as follows: i) stability controller; ii) gain
Figure 2 – Single line diagram of TMMI optimizer; iii) undervoltage strategy; iv) overvoltage
strategy; among others.
2. Considerations on TMMI SVCs
2.2. Basic characteristics of TMMI FSCs
and FSCs
It is worth noting that manufacturer I also supplied 18
2.1. Basic characteristics of TMMI SVCs FSCs that compensated the 500 kV TMMI transmission
The main characteristics of TMMI SVCs are as follows: lines by 70% on average. The FSC locations are shown
in Figure 2. All TMMI FSCs are equipped with MOVs
i. Jurupari SVC – 2 TCRs rated at 149.6 Mvar each; and triggered spark gaps, as shown in Figure 4. The
2 TSCs rated at 134.6 Mvar each; 2 single tuned FSCs are rated at 35% of the lines reactance, totaling
filters (5th harmonic) at 20.6 Mvar each – 60 Hz
70% per line, except the FSCs installed on the 500 kV
and 18.3 kV (secondary voltage). Output -200 to +
Tucuruí-Xingu lines (circuits #1 and #2), at Xingu side,
200 Mvar on 500 kV, 13.2% coupling transformer
which are rated at 70% of the lines reactance.
reactance (200 MVA).
ii. Oriximiná SVC – 2 TCRs rated at 179.5 Mvar each;
2 TSCs rated at 212.5 Mvar each; 2 single tuned
filters (5th harmonic) at 23.9 Mvar each – 60 Hz
and 15.5 kV(secondary voltage). Output -200 to +
300 Mvar on 500 kV, 15% coupling transformer
reactance (300 MVA).
iii. Macapá SVC – 1 TCRs rated at 152.1 Mvar each;
1 TSCs rated at 190.6 Mvar each; 2 single tuned
filters (5th harmonic) at 25.7 Mvar each – 60 Hz
and 15.5 kV (secondary voltage). Output -100 to
+ 160 Mvar on 230 kV, 15% coupling transformer
reactance (160 MVA). Figure 4 – Simplified single line diagrams of TMMI FSCs
2.3. General system requirements specified in vii. It shall be demonstrated the performance of the SVC
auction’s technical annex for operation in normal and degraded conditions
through digital studies, such as electromagnetic
It is important to highlight that some system requirements
transients studies, and both transient and dynamic
have been specified in the Technical Annex related to
stability studies, to be conducted by the transmission
Lots A, B and C’s auctions. Those requirements have
entity.
ensured that additional studies had to be conducted to
better understand and enhance the SVCs performance 3. Series-compensated AC
under system disturbances. The most important system
requirements are listed as follows. systems and SVCs – Possible
i. The SVC control systems shall not compromise adverse interactions
the BIPS performance, either in normal operation
or under contingencies, emergencies and degraded 3.1. An overview on main resonances in an AC
operation, for steady state and transient conditions. series-compensated power system
ii. The SVCs shall not give rise to ferroresonance The series compensation by fixed series capacitors in extra
conditions or asymmetric core saturation. high voltage systems has been recognized as an efficient
and powerful means to improve transient stability and
iii. The SVC control shall be designed to prevent thereby increasing the transmission capacity between the
hunting with controls from other electrically near power systems, particularly in inter-regional AC links.
SVCs. The switching operations of SVC branches However, fixed series capacitors cause subsynchronous
or external elements, ordered by the control system and supersynchronous oscillations in power systems due
including also on-load tap changers of transformers, to several resonances in power systems.
shall not lead to intermittent oscillations (hunting)
among these elements, nor between any of them Figure 5 shows two series-compensated lines in a typical
and the SVC control. series-shunt-compensated power system.
iv. The SVC control system shall be designed in order The part of a series-shunt-compensated power system
to help minimize the disturbances in the electrical shown in Figure 5 is useful for illustrating the main
system for a fault. The control system shall be resonances [2], [3], namely;
designed considering the need for single-phase i. Line shunt resonance (dotted black rectangles) – this
autoreclosure without tripping the SVCs. type of resonance is well known and low frequency
v. The SVC control shall take into account additional but yet above fundamental (synchronous) frequency.
quantities (power flow, reactive power flow, etc.) It comes from a parallel resonance between series
with the aim of phase-modulating (if necessary) inductances and shunt capacitances of transmission
the SVC output towards dampening voltage lines.
oscillations, power oscillations in the power grid ii. Series resonance (dotted blue rectangle) – this
and subsynchronous resonances. resonance is due to the combined effect of series
vi. The SVC control shall be designed in order not to inductances of lines, series capacitors and generators.
affect the stability of the grid voltage. This shall The frequency of this resonance increases with the
be done by identifying the sensitivity of the grid degree of series compensation. This series resonance
voltage to the variation of the SVC susceptance, causes subsynchronous components that may impact
and taking corrective action to avoid instability adversely turbo-generators. These phenomena are
conditions. named subsynchronous resonance (SSR).
flashes in a 10 second-period. The gain reduction scheme Once the SVC controller identifies SSOs on its voltage,
can be summarized in Table I. it starts checking the condition for enabling the SSO
Table I – Gain reduction scheme for SVCs on 500 kV system function. Note that the first gain reduction (RELATIVE_
GAIN_VREG) is caused by the stability controller
Number of SVCs in nearby operation in 1 2 3
Voltage Control Mode (0.70 to 0.44 pu) and an oscillation around 35 Hz is
seen by the voltage regulator (VRESP). Therefore, as
Gain factor applied to Jurupari and Oriximiná 1.00 0.60 0.36 soon as the triggering conditions of the SSO function
SVC controllers (pu)
are met, the relative gain is temporally reduced from
Gain factor applied to Silves SVC controller (pu) 1.00 k1 k2 0.44 pu to pratically zero, freezing the SVC controller
remark: k2=k12 and k1 is a function of the short circuit power at 500 kV as a consequence. At the time that the SSO signal falls
Silves busbar below a set threshold, the Silves SVC’s relative gain is
ramped back within 500 ms to its original value set by
7. TMMI operational experience - the stability controller (0.44 pu), enabling the Silves
Contingencies occured on TMMI SVC to operate back in voltage control mode. During
this return, the SSO signal is still monitored and if
It is worth remembering that TMMI can supply up to it becomes higher than the threshold level, the SSO
50% of the Manaus and Macapá load in order to ensure function is resumed, i.e. the voltage regulator is frozen
that the underfrequency load-shedding scheme is able once more.
to recover the load-generation balance of both Manaus
and Macapá systems. This requirement is associated to Figure 9 depicts the Oriximiná SVC controller behavior
the system dynamic performance regarding the balance during a single-phase-to-ground fault on the 500 kV
between load and generation during system islanding Oriximiná-Jurupari line (circuit #1), which led to the
after multiple contingencies. activation of the new digital filtering. An oscillation
around 12 Hz can be measured by the SVC voltage
Since the commissioning of TMMI SVCs, there have regulator in the VACT signal. Therefore, as long as the
been a large number of disturbances such as line outages, conditions to enable this new digital filtering are met,
internally and externally to this interconnection, resulting Oriximiná SVC gain is reduced by 50% and it starts
sometimes in islanding of Manaus and Macapá systems. running on this feature within the time frame of the
Figure 8 depicts Silves SVC performance during a oscillography.
system disturbance (severe fault outside TMMI).
KEYWORDS
Power system stability, power system control, power system protection, power system dynamics, power plant
protection, transient stability
“Out-of-phase” is almost identical to “out-of-step”. quickly reconnect once the power grid conditions have
However “out-of-step” includes a pole slip of one stabilised.
or more generators, while “out-of-phase” does not
1.1 System perspective
necessarily require an out-of-phase angle larger than
180°. The term “out-of-phase” is typically used by Major disturbances in power systems are engraved in
circuit breaker experts for whom switching under out- the societal mind as their effects are unacceptable to
of-phase conditions covers conditions with an out-of- modern societies. With a failing power system, society
phase angle of, say, 90° and above. “Out-of-step”, used quickly stops functioning. The larger the black-out
by power plant experts, protection experts and system area and the longer the period without electricity is,
study experts, covers a particular case of “out-of-phase”. the larger the societal chaos and economic damage will
be. The economic losses outweigh by far the costs of
Power plants have to be protected from severe damage
the electricity involved and the costs of the necessary
and should therefore be equipped with fast acting
improvements in the local infrastructure. Resilience is
protection devices in order to detect faults inside the
a major topic nowadays on the political agenda. The
power plants (generators, transformers, auxiliary
scenarios described in this article are failures due to
systems and relevant connection buses and cables). From
natural catastrophes, human activities and also due to
a power system point of view, it is essential that multiple
the infrastructural system itself. The requirements for
power plants do not trip at the same moment and the
a power system to prevent large disturbances due to
number of power plants tripping for a single disturbance
system instability, or to reduce its impact, are addressed.
is minimised. Therefore, protection equipment should
discriminate between disturbances within the power A blackout may be caused by a lack of power and/
plant that may damage the power plant itself, and or reactive power in relation to the load or by a lack of
disturbances on the power system that may damage the power transfer capacity. In the first case, the local network
system’s integrity, but can be withstood by the power will show serious problems of frequency, while voltage
plant for a limited amount of time. If tripping of power instability might occur in the second case before collapsing.
plants (in the MW range or larger [1]) is unavoidable, it In the latter case, the electrical power transfer will become
is also requested that they keep running while isolated too large which violates angular and voltage stability as
from the main power network and should be capable to well as thermal limitations of the transmission corridors
protection). In the latter case the total clearing time This requirement can be expressed as a minimum value
may reach several hundreds of ms (fig.3) and selective of the CCT or as an under voltage ride through (Fault
clearing of the faults is not assured. Ride Through: FRT) curve. FRT requirements are set
forward either in (inter)national grid-codes or by TSOs,
Figure 3 is a general example of the main and back-up
or are negotiated between grid and plant operators.
protection fault clearing time. Often, the back-up clearing
Figure 4 extracts FRT-requirements from the recently
time of the grid protection is longer than in figure 3 in
approved European grid-code “Requirements for
order to avoid the risk of non-selective tripping. Its setting
Generators” [1]. TSOs can determine the precise values
depends on the characteristics of the network (including
between the minimum and maximum limits given by the
the automatic re-closure philosophy and timing) and the
code. Note that the requirements for non-synchronous
protection system (delay in telecommunication signals,
generators connected with inverters are higher than those
relay reset time, etc.). In addition, slower protection
for synchronous generators, as they can easily fulfil the
functions (third and fourth zone of distance, (directional-)
longer duration of voltage recovery.
overcurrent, etc.) are installed as back-up protection
functions for clearing the fault if the main protection fails
or clearing far reach faults.
Power plant protection should not trip on a grid fault
before the grid main and back-up protection clearing
time. Some TSOs further request that the generator
protection trips are even more delayed (for example in
order to be selective with the third zone of the distance
protection). A medium to large synchronous generator
generally contributes short circuit current in the range
of 2 to 4 p.u. The IEC [45] and IEEE Standards [46]
state that generators should withstand these fault
current levels for 1 s or more. A similar consideration is
applicable for the generator withstand capability to the
negative sequence fault current components caused by Figure 4 - FRT limits to the European Requirements for Generators
non-symmetrical faults in the grid. larger than 50 MW Minimum and maximum requirements are given for
synchronous generators and non-synchronous (PE)
To prevent power plants from tripping due to angular
Apart from such a coordination, some power plant
instability the grid protection must operate faster
protection should unambiguously detect faults within
than the power plant’s critical clearing time [1]. For
the plants’ territory, in particular differential protection
strategic parts of the network, back-up protection
(generator 87G*, step-up transformer 87T*, step-up
should act faster than the CCT in order to avoid angular
transformer earth restraint 87N*), stator earth fault
instability and disconnection of the power plant. When protection (59G* or 51G*), loss of field protection
the CCT is around or below 200 to 300 ms, it might be (40 G*), etc. Although protection cannot prevent the
challenging to respect this principle [33]. In such cases, dynamic stresses of the first peak of a sub-transient
applying a redundant instantaneous protection system short-circuit current, the cumulative damage caused
can be considered such as duplicate main protections by succeeding current peaks can be limited. When a
(differential protection or distance protection with tele- circuit breaker is closed under out-of-phase conditions
protection and with independent communication links). - for instance at false synchronization - the out-of-
Based on fault clearing times and the recovery of the phase currents may exceed the peak value of the sub-
voltage after fault clearing, TSOs request generators to ________________
withstand a voltage dip curve without losing stability. *ANSI-codes for protection functions; see figure 5 and [11]
transient short-circuit current [34]. According to the At power plants an out-of-phase condition may occur
international standards for large generators [45][46], in in case of mistakes in the wiring or the setting of
case of a generator terminal short-circuit or an out-of- synchronization apparatus. Error in the connected
phase synchronizing occurrence, the generator stator phases, including wrong rotation direction, leads to
windings, rotor windings, generator and turbine shaft, false synchronization with a large difference in phase
shaft couplings, shaft balance and foundation should angle between generator and network. This can happen
be checked carefully, in order to prevent consequential at the first synchronization of a new generator or after
damage in service. important modifications of the power plant and/or the
Table I - Summarized overview of enquiry results among TSOs worldwide (1 to Z) on their power swing protection policy
TSO 1 2 3 4 5 6 7 8 9 A B C D E F G H I J K L M N O P Q R S T U V W X Z
predefined
N N N N N N N N N N N N N N N N Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
locations1
out-o-step
N N N N N N N N N N N N N N N N Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
function2
special
N Y Y Y Y Y Y Y Y Y Y Y Y Y Y
function3
blocking
N N N N N Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y Y
distance4
which
2-4 1,2 2-4 all 2-4 all all all 1 all 1 1,2 1,2 1,2 1,2 1,2 all all 1,2 1 1 1 1 1 1 1-3 1,2 1,2
zones
1 Whether TSO applies pre-defined locations for system separation in case of unstable power swings
2 Whether TSO applies out-of-step relays ANSI 87 or out-of-step tripping function in distance relays ANSI 21
3 Whether TSO applies special protection systems, such as voltage swing relays, inter-trips, voltage phase angle comparison, line loading.
The application of centralized wide area protection systems has not been reported.
4 Whether distance protection relays are blocked for unstable power swings and which zones are blocked
conditions of TEPCO, Japan, where a tight power areas that are coupled by HVDC links. These areas may
generation capacity margin has arisen since nuclear support each other through the HVDC connection with
power plants were taken out-of-service. By using a risk that the neighboring area will also lose stability or
diurnal energy storage (filling the water basins during HVDC connections can be blocked, thus restricting the
the night and releasing the water during the day), energy incident to one synchronous area only.
demands can be met, but the loss of a transmission line Not much has been published on the reliability of the
to the hydro plants would lead to stability problems. existing and the new (proposed) protection schemes or
Therefore, an automatic hydro plant shedding scheme on the reliability of distance protection with blinders
is being implemented to keep the system stable. This or other software tools to detect very slow or very fast
is a new special protection scheme, which will operate power swings. In Ireland and Denmark power system
autonomously at each hydro plant without the need of oscillations with a frequency of only 0.4 Hz occurred. In
analyzing remote information. See for other applications the UK the Grid-code requests that the generator power
with respect to the state of the art [12][17]. system stabilizers are tested in order to prove damping in
According to the Russian grid-code [2] the control the range from 0.3 to 2.0 Hz. Note that fast oscillations
centre intervenes manually by load shedding when long between power plants could be in the range of 4 to 7 Hz
interconnecting overhead-lines are loaded above 80% (and even more: in Russia 10 Hz and in [36] 11 Hz).
of their secure power transfer stability limit, as defined Close collaboration between the engineers who perform
by the TSO. Apart from that, system wide contingency the stability studies and the protection engineers will
analyses based on a state estimator are used to detect contribute towards the development of good settings.
possible stability problems and simulate the effect of load Power swings are complex events. While evaluation
shedding. Then, load shedding follows automatically by of relay settings and configuration in simulations will
the same technology as applied for under-frequency load contribute to the development of secure and reliable
shedding. If automatic load shedding is not effective and settings, additional understanding and confidence in
the power system falls into out-of-step conditions, the relay algorithm and setting performance can be gained
network is split by special out-of-step tripping relays at from lab testing or commissioning testing of the
pre-determined locations. relay using simulated measurements. These may be
The described system protection schemes are examples generated from offline studies using conventional RMS
of schemes used to prevent system splitting, but there are stability assessment programs (PSSE, PSLF, TSAT,
also protection systems that act in order to keep parts of the PowerFactory etc) or from on-line studies using real-
system stable after system separation. An example of this time digital simulators (RTDS).
is where tie lines between regions are tripped to preserve An extensive type test of the algorithm to detect power
the stable operation of the grid within each region. In swings and to block or to give a (delayed) trip is
such cases, the regions do not support each other any performed in Russia [39]. A network is used, as given
longer. This is a matter of policy: the grids support each in figure 8 and the relays locate at the buses 2 and 5. A
other so long as practical and when appropriate separate number of cases with different load, faults and power
the healthy parts to safeguard them. A similar choice in swings are simulated with an RTDS simulator and the
policy could be made between autonomous synchronous behavior of the protection algorithm can be extensively
tested. But it is not possible to simulate the cascading a large retrofit program has been scheduled and it has
effects that occur in practice. been planned to be implemented before 2016, at least in
Germany and in Italy.
4. Dispersed energy resources In smaller synchronous areas, like on the Island of Ireland,
(DER) affecting power system the active power frequency response of renewable
stability power sources is very critical. This resulted in specific
requirements in the grid-code [3], illustrated in figure 10.
Conventional synchronous power plants control the system
frequency in such a way that deviations from the nominal
frequency are minimized through being immediately
compensated by (1) the inertial energy stored in the
rotating masses, (2) the primary governor response (droop
control), and (3) secondary and tertiary control regimes.
Nowadays frequency deviations occur more often, with
a longer duration and with a larger magnitude than in the
past. A main cause is varying energy exchanges forced by
the market prices that are adapted for each hour. There are
some disputes about the influence of increasing percentage
of DER. The renewable sources are volatile. Thus, by
their large numbers, could enlarge power frequency
deviations by certain changes in weather conditions,
unless suitable reserves are available. Nevertheless, more
relevant is their response to frequency deviations – before
relevant grid-code changes were (are) enacted they were
(are) permitted to trip for modest deviations of frequency.
Figure 9 extracted from an ENTSO-e report [18] shows Figure 10 - Irish gridcode requirements for active power frequency response
of windfarms [3]
that the large amounts of DER tripped before 49.5 Hz
resp. 50.5 Hz would be reached, while the European grid- At a major disturbance, it could happen that some power
code [1] requires operation at least between 49 and 51 Hz plants do not act as expected, and/or load shedding do not
for all power generating units. Meanwhile, the quantity of follow the rules that predetermined percentage of load
inverter-coupled generation found to have disconnected will be dropped immediately at a certain under-frequency.
in figure 9 exceeds 3000 MW, the level of a sudden power It is evident that the escalation of an incident could be
deficit up to which the European continental system has avoided by applying and verifying the strict rules which
been designed to control the frequency. have been agreed upon by the collaborating parties.
During the system split of the European network on 4 Another challenge is that certain types of renewable power
November 2006, a large number of wind power parks generation lack synchronous inertia and synchronous
not only disconnected automatically at a certain power torque [40]. In case that conventional generation is
frequency deviation, but also reconnected automatically replaced, an increasing share of inverter-interfaced DER
at a certain smaller frequency deviation. Figure 9 shows leads to a decrease of system moment of inertia necessary
the amount of prospective disconnected DER capacity in to achieve system stability. However, the situation is more
case of frequency deviations [18]. Due to this assessment, complex than at first glance, as conventional plants are
• Unlike the USA, which has National Electric Reliability [4] Qikai Zhuang, “Managing Risks in Electrical Infrastructure
Assets from a Strategic Perspective”, PhD thesis Delft University of
Standards per the North-American Electric Reliability Technology, 2015, ISBN 978-94-62331-25-9
Corporation (NERC), Europe does not have over-
arching rules on the required response of generator and
Technical Reports:
grid protection to power swings. Instead the details are
[5] IEEE PSRC WG D6 Report to the Power System Relaying
left open to the TSOs in their respective grid-codes, Committee of the IEEE Power Engineering Society, “Power Swing
even where TSOs are connected within the same and Out-of-Step Considerations on Transmission Lines”, 2005-07-
synchronous area. It is technically beneficial to have a 19
unique, unambiguous set of rules within a synchronous [6] ENTSO-E/TEIAŞ publication, “Report on Blackout in Turkey on
area. Such a ruling should not depend so much on 31st March 2015”, www.entsoe.eu, 2015-09-21
each TSO, but on uniform system wide and local [7] CIGRE Technical Brochure 316 (2007), “Defense Plan against
technical aspects, in order to achieve fairly distributed Extreme Contingencies”
responsibilities. Such rules can even cross synchronous [8] UCTE publication, “UCTE Final Report System Disturbance on 4
areas, depending on the way HVDC couplings are November 2006”, www.entsoe.eu, 2007-01-30
operated under transient phenomena. [9] CIGRE Technical Brochure 613 (2015), “Protection of Distribution
Systems with Distributed Energy Resources”
• In the international (generator) circuit breaker
[10] Deutsche Energie-Agentur (dena), Berlin, “Ancillary Services
standards, requirements for type testing under out-of- Study 2030”, www.dena.de,/en
phase conditions have been defined. Other prevailing [11] CIGRE Technical Brochure 479 (2011), “International Guide on
conditions, such as power frequency voltage deviations the Protection of Synchronous Generators”
from rated values and actual delay times in protection [12] CIGRE Technical Brochure 187 (2001), “System Protection
and auto-reclosing settings, have to be considered in Schemes in Power Systems”
project specific applications. For instance, whether and [13] CIGRE Technical Brochure 608 (2015), “Lessons learnt from
which protection takes care for immediate tripping after Recent Emergencies and Blackout Incidents”
false synchronization or the relationship between out- [14] Electra No. 50, 1977, pp. 77-92, WG 34.01, «Protection against
of-phase angle and the actual operating voltage. Out-of-Step Operation of Large Synchronous Machines»
*simon.gill@strath.ac.uk
KEYWORDS
Power System Security, Reliability, Security of Supply, Transmission Adequacy, Wind Generation.
parties reflecting their contribution. When comparing the Table I: breakdown of main generation plant in Scotland
as of December 2015.
level of imports required to the transmission networks
_____________________________________________________________
import capability, care should be taken to model the
network, the initial generation dispatch and outages Generation Main Installed Notes
Type Stations Capacity
appropriately as power is rarely shared equally between (MW)
________________________________________________________________
circuits on a region’s boundary and the critical outage or
limiting factor may be remote from that boundary. The Due to Close
Coal Longannet 2400
March 2016
deterministic rules in the SQSS relating to conditions at _____________________________________________________________
time of peak demand address either a relatively modest Operational capacity
Gas Peterhead CCGT 740
transfer with a double circuit outage, a single fault in 2015/16
_____________________________________________________________
outages on a system that already has one other network
Hunterston B 1016
outage, or a higher transfer with a single outage. The Nuclear
first two condition are typically the more limiting ones. Torness 1280
_____________________________________________________________
Thus, in this paper the term ‘secure capability’ refers to Pumped Cruachan 440
N-2 security, that is, secure after a double circuit outage storage Foyers 300
or one fault following some other prior outage. _____________________________________________________________
31 Locations with 39
This paper uses the example of Scotland as a region of BM wind a
Balancing Units 3142 As of Autumn 2015
the GB power system to consider the need for secure _____________________________________________________________
import capability in order to secure demand in the Other wind Distributed 2170
region. The impact of major plant closures and the role _____________________________________________________________
of the existing and an expanded Scottish wind fleet are BM hydro a 896
_____________________________________________________________
investigated, before discussing how the results suggest
a
BM generators are those that are active in the balancing mechanism with data on
various parties contribute to regional security of supply. availability at half hour resolution, it includes a mixture of transmission and distribution
connected generators. ‘Other’ generation is not monitored, and its output appears as a
it is relatively sparsely populated. In January 2016, curves are shown in Figure 2 (b): the situation if wind
approximately 5GW of wind generation was operational generation had been allowed to run un-constrained,
in Scotland, with a further 1.6GW in construction and a and the situation if all Scottish generation had run
further 6.8GW consented [14]. unconstrained – i.e. if the SO had taken no actions to
Scotland is connected to the rest of GB via two 400kV keep export levels within limits. Such unconstrained
double circuits (and some minor 132kV circuits) operation would require significantly greater export
crossing the ‘B6 Boundary’. Transmission boundary capability than is currently provided, with exports rising
capabilities are listed annually in National Grid’s to greater than 5GW during some periods.
Electricity Ten Year Statement, and the value quoted
for export from Scotland in 2016/17 is 3.5GW [15];
3. Security of supply in Scotland
the import capability is 2.6GW [16]. New transmission Security of supply calculations attempt to strike a balance
infrastructure is expected to be commissioned in 2017 between the risk of needing to disconnect customers and
increasing both the import and export capability. the cost of providing the capability to meet demand. In
Replacement of conductors on circuits within Scotland
this paper we calculate an estimate of the secure import
in the near future will increase the export capacity to
capability that would be needed in order to ensure that
4.4GW; and a 2.2GW, 600kV embedded HVDC link
demand in Scotland is met for all but a certain number
from Hunterston in central Scotland to Deeside in North
of hours within a winter. This definition is similar to
East Wales will increase the secure export capability to
that of LoLE used in generation adequacy studies. To
6.6GW and the secure import capability to 3.8GW [15].
differentiate the transmission adequacy calculation we
Figure 2 shows cumulative distributions for demand in call this the ‘Regional Demand Reduction Expectation’
Scotland [17], and estimated boundary flows for 2015 (RDRE) requirement. To illustrate the method we study
based on nominated outputs from generators in Scotland two levels of security: an RDRE requirement of no more
at gate closure known as Final Physical Notifications (1 than 3 hours per winter (in line with the current GB
hour ahead of delivery) and post-gate closure balancing generation adequacy standard), and a stricter limit of 1
actions taken by the SO [18]. Peak demand in Scotland hour per winter. Four scenarios are investigated:
over the winter of 2014/15 was approximately 5.5GW
[19]. 1. Base case: the intact system during the winter of
2015/16 including Longannet power station.
Figure 2 (b) shows that Scotland currently exported 2. Longannet Closed: Longannet power station closed,
the vast majority of the time: 93% across the full year representing the system after March 2016.
of 2015. The two nuclear power stations operated as 3. Low conventional: Longannet, Peterhead and
baseload when technically available, and generation at Hunterston closed.
Longannet coal power station was used the majority 4. High wind future: Longannet, Peterhead and
of the time. Periods of import were mainly during day Hunterston closed and wind scaled up to represent an
time, concurrent with summer maintenance outages at additional 5 GW of capacity.
the main power stations and coinciding with low wind
availability. In each case, demand in Scotland is distributed as it
was over the two winters of 2013/14 and 2014/15.
Estimated historic exports from Scotland in 2015
Generators including hydro and pumped storage are
reached a maximum of 3.5GW consistent with the secure
modelled at their 2015 capacities, and for scenarios 1
export limit reported in [16]. However, ensuring that the
– 3 two sensitivities are run to identify the role of the
boundary flows remain within that limit requires action
existing wind fleet on security of supply:
by the SO to constrain down generation within Scotland
when flows would otherwise exceed it. Two additional a. Existing large Wind Fleet: uses historic data from
Figure 5: (a) cumulative and (b) density plots of convolved and time series estimations (denoted C and TS respectively)
of boundary import requirements for the base case scenario with and without wind.
Figure 6: cumulative distribution for ‘Longannet Closed’ and ‘Low Conventional’ cases estimated by convolved and time series analysis
(denoted C and TS respectively) with and without the existing large wind fleet
The sign convention of each distributions is chosen so thought of as a market neutral median power flow at time
that the need for import into Scotland is represented by of system peak demand, is from Scotland to England).
a positive number. For negative ‘import’ values there However, a small import capability is required under
is in effect zero necessity for import into Scotland the RDRE calculations if wind generation is ignored.
(Although the market may choose to import under some With wind included the RDRE method agrees with the
circumstances, it is not required in those conditions SQSS. Whilst in this case the difference is marginal, it is
for security reasons. Negative values reflect the secure likely that in future it will not be appropriate to designate
boundary capability required in order to allow all regions as purely importing or exporting; with high wind
available Scottish generation to be used - maximum penetration and low conventional capacity in a region,
unconstrained export). it will be important to define both import and export
requirements for each region.
Results show that there is very limited requirement for
import capability in the base case. Import is required for Whilst the SQSS explicitly ignores wind capacity in
security reasons during only 1.6% of the winter when the security calculation, the impact of wind on the
if wind is not considered, or 0.02% of the winter when distribution of boundary flows can be seen in the RDRE
wind is considered. values of Table II. With Longannet closed, results with
and without wind show that the 2.5GW of wind capacity
The closure of conventional plant will increase the need for added between the two sensitivities reduces the required
import. Figure 6 shows results for the two scenarios with import capability by 630MW and 590MW for the 3
station closures. The closure of Longannet increases the hour and 1 hour cases respectively; this is, in effect, a
fraction of time that Scotland is likely to require imports ‘capacity credit’ of approximately 25%. Similar results
for security reasons from 1.6% to 10.6% without wind and are seen in the Low Conventional scenario, although the
from 0.2% to 1.8% with wind. The closure of Peterhead impact of the 2.5GW wind capacity is slightly reduced.
and Hunterston further increases the fraction of time import The addition of an extra 5GW of wind in the High Wind
is needed to 69% without wind and 24% with wind. Future scenario further reduces the required import
capability, however the impact is significantly smaller
6. Defining a secure import and the reduction in secure import requirements due to
capability requirement this extra generation is only 5% of its capacity.
The secure import capability required for a particular The reason for the relatively high impact of the first
RDRE requirements can be calculated from the 2.5GW of wind is that while the availabilities of wind
distributions shown above. The 1 and 3 hours RDRE within Scotland is strongly correlated across wind farm
requirements used in this paper correspond to finding sites at any one time due to weather conditions across the
the P99.90 and P99.96 value of the boundary flow region. Wind availability, as a whole, is not correlated
distribution. Table II shows estimated secure import with the availability of conventional units (and therefore
capabilities for the scenarios and sensitivities studied provides a source of generation-availability independent
calculated against the RDRE requirement and compares of other sources) but new wind farms added to an
these to values calculated using the current SQSS existing wind fleet are correlated to existing generation
deterministic methodology. and therefor have a smaller effect on secure import
requirements. In this respect new wind added to an
The results highlight several important issues. In the existing wind fleet contrasts with new conventional
base case the SQSS does not require a secure import units: an additional conventional unit provides a new
capability because Scotland is defined as an exporting independent source of generation availability, whilst an
region. (The ‘planned transfer’, which can be best additional wind farm does not.
____________________________
2- To be more precise, the SQSS defines two ‘planned transfers’: one intended to reflect ‘economy driven’ need and the other ‘security driven’ need. Depending on
the ‘generation background’, it is possible that these would point in opposite directions.
* ivan.dudurych@eirgrid.com
KEYWORDS
Control centre tools, Non-conventional generation, Operational Security, Stability, Power transmission System
Steady-‐state Dynamic
oscillations. Oscillatory stability encompasses small will be significantly different from day to day and from
signal stability, as well as transient and frequency stability hour to hour, causing it to change outputs and even start/
where disturbances could result in unstable oscillations stop much more often than in the past. Figure 3 shows
of active and reactive powers, frequencies and voltages. changes in the scheduling of one combined-cycle gas
Figure 1 shows the structure of operational security turbine generator (CCGT) in Ireland between November
components discussed above. 2009 and November 2015. The scheduling changes for
this particular CCGT are due to the combined effect of
The components of the operational security and its criteria
increasing the renewable generation, connection of a new
are described in [2].
CCGT, and cheaper coal prices.
3. Impact of very high penetration of With the advent of renewable and other distributed
generation, transmission systems became subject to a
renewable sources on operational much greater variety of power flows, while distribution
security systems became subject to multi-directional power flows
depending on the prevailing generation conditions. Figure
3.1. The Effects of Resource Variability
4 shows two EMS screenshots displaying power flows
Renewable resource variability has a profound impact at some transmission substations, which have connected
on operational security. Figure 2 shows wind generation wind generation. Not only can the level of power flow
in the all island (Ireland and Northern Ireland) power vary, but also the direction of the flow: with low wind,
system as a percentage of the total generation for October- power flows from the transmission to the distribution
November 2015. It is seen that such generation is quite system, while with high wind it can change its direction
variable and changes without pattern from almost zero to as seen in Figure 4 (right) and net power flow can be from
55%. This means that scheduling of the conventional plant the distribution to the transmission system.
Figure 4 Power flows at the transmission substations with connected wind generation, with low (left), and high (right) wind conditions: Red arrows
indicate net flow from the transmission to the distribution system, and green arrows indicate net flow in the opposite direction
that voltage dips associated with faults (three- or even two- compromising operations of the distribution system. This
phase) on the transmission system propagate to a broader is particularly relevant given that about a half of the total
region of the system. This has a direct impact on the fault- wind capacity is connected at the distribution level.
ride-through performance of non-synchronous sources
3.4. Effect of renewables on frequency stability
of generation. The larger the propagation of the voltage
dip into the system, the greater the proportion of non- The main effect of non-synchronous plant on frequency
synchronous generation that may be exposed to an under- stability is related to the fact that renewable generation
voltage protection trip. In addition, unlike conventional provides very little or no inertia to the system, because
generation, some renewable generation, in particular wind it is partially or completely mechanically decoupled
generation, has a slower rate of recovery of active power from the power system. The system inertia is getting
following the clearance of the fault as illustrated in Figure 6. smaller and smaller with progressive replacement of the
This can lead to a new phenomenon – Voltage Dip Induced conventional generation by non-synchronous renewable
Frequency Dip (VDIFD) – and operational security should generation. This fact is clearly demonstrated at Figure 7
take account of the probability of such an event. where inertia is represented as kinetic energy stored in the
rotating masses of the synchronous machines.
3.3. Effect of renewables on voltage stability This means that a sudden imbalance of power will cause a
In the early days of renewable energy, renewable generators greater frequency deviation in a system with less available
did not have any voltage or reactive power control inertia, as shown in the illustrative Figure 8. The curve
capability. Soon it was recognised that the availability of (c) in this figure corresponds to a situation where the
voltage control capabilities of the conventional generators load-shedding scheme would activate. Load shedding,
was increasingly depleted when more of them were the automatic disconnection of load, is a last defence
replaced by renewable ones. The Grid Code was updated protection and its routine operation is not accepted in a
to introduce at least some of the capability requirements modern society.
of conventional generation for the non-synchronous From the above analysis, it follows that we need to
generation portfolio. For instance, the Irish Grid Code monitor the inertia and frequency stability in real time.
requires that all transmission-connected wind farms have
the ability to control voltage and/or reactive power at
their terminals. The full-rated converter grid-connection
technology is best suited for this task as it allows
modulation of the grid-side three-phase voltage according
to the required reactive power. On the other hand, many
distribution-connected renewable sources on the Ireland
and Northern Ireland system are fixed power factor
machines and they do not regulate voltage. This means that
voltage regulation becomes increasingly more difficult in
a system with high penetration of renewables. In areas
with significant renewable generation, network solutions
are required to avoid voltage collapse or excessively high
voltages. It is important for the Distribution Code to be
updated to obtain the reactive power capability from those Figure 7 Decline of system Inertia with increasing of wind penetration as %
wind farms connected to the distribution system without of generation
HVDC Interconnector Imports are included here as they • synchronous torque for system strength and to maintain
displace the conventional generation together with other transient stability, and
non-synchronous sources. At a constant demand, an • reactive power support (provided there is a suitable
increase in non-synchronous generation would equate geographical spread of generation) to manage voltage
to an increase in SNSP. Maintaining SNSP below a control and stability.
defined value ensures the robust operation of the system
Increasing nonsynchronous generation places pressure
at differing levels of non-synchronous generation. The
on existing system inertia limits. A balance must be
existing maximum SNSP limit is set as 60% on a trial
maintained between levels of synchronous and non-
basis [1]. The maximum SNSP limit is maintained by
synchronous generation. Lowering the system inertia limit
curtailment of non-synchronous renewable generation.
would permit a reduction in the quantity of conventional
System inertia. Monitoring system inertia from generators synchronised to the system. Maintaining the
synchronous generation guarantees a minimum level of existing inertia requirement level would at times increase
conventional generation is synchronised to the power curtailment of non-synchronous sources on the system.
system. At present this minimum level of inertia is set for
RoCoF. Rate of change of frequency (RoCoF) is the
the Ireland and Northern Ireland system as 20,000 MWs
measure of frequency over time following the sudden
[1]. System inertia is also maintained through an alternative
disconnection of a generator or an interconnector.
operational metric, which ensures a minimum number of
Predominantly, the most vulnerable time for RoCoF to
generating units are synchronised on the system. This
exceed its limit is in the initial moments after the event. To
minimum level of conventional generation is set to provide:
this end, Equation (2) is a technique to calculate RoCoF
• inertia to manage frequency control and stability, and therefore allow monitoring of this change.
5. Control centre tools Through the tools’ user interface (see Figure 9), the grid
controllers specify the maximum wind production for the
system or a group of windfarms to ensure operational
5.1. Monitoring the new operational metrics security. The tool selects windfarms to be dispatched
The new operational metrics discussed above have been down and calculates setpoints based on a regulatory
included in the generation scheduling tool to ensure defined hierarchy. Such automation is essential for
that operationally secure plans are created and allowing efficient decision making in complex real-time situations.
decisions to be taken in a timely fashion e.g. day ahead. Constraint and curtailment can co-exist - for example,
For example, additional synchronous generation may be a thermal overload contingency may require local wind
required for inertia – optimising this in advance allows generation to be dispatched down while minimum inertia
the transmission system operator (TSO) take more requirements may require concurrent curtailment to avoid
economic actions. It also allows trading of displaced wind having to decommit a conventional unit. When removing
generation on HVDC interconnectors to Great Britain, curtailment, the tool respects active constraint setpoints to
thereby reducing curtailment. avoid exacerbating any local security issue.
Grid Controllers monitor these metrics on-line via
the Energy Management System (EMS), to cater for
unexpected scenarios such as higher than forecast wind
generation or reduced system inertia due to the loss of a
large synchronous generator. We implemented specific
calculations and displays, which show the prospective
RoCoF for the loss of any one of the large synchronous
generators or HVDC interconnectors. To allow pre-
emptive action, grid controllers are warned of situations
approaching operational limits (e.g. 0.45Hz/s RoCoF)
and second level alarms activate once limits have been
breached (e.g. 0.5Hz/s RoCoF).
In a number of cases, the generation dispatch has been enhanced power system analysis functionality based on
changed to remove frequency insecurity risk (e.g. a large an all-island network model. One of the main advantages
frequency deviation following a trip of a unit or following of the integrated system is the operational ability to fulfil
a system separation event between Ireland and Northern more complex real-time functions across two jurisdictions.
Ireland). Specific applications facilitate easier scaling of numerous
wind generators in powerflow studies and improve
In July 2015, we started providing the Grid Controllers
monitoring of negative reserve provision (capability
with real-time information (refreshed every 5 minutes)
to reduce generation to prevent over-frequency) and
about the largest frequency nadir and zenith that can longer term reserve provision (e.g. ramping capability to
arise from an N-1 contingency, and from July 2016 we balance a renewable forecast error). Improved situational
also included information about maximum and minimum awareness and all-island wind dispatch capability are
RoCoF values as shown in Figure 13. other features.
As a new set of system services products is being
5.5. Integrated EMS
introduced [6], some of them (e.g. ramping products,
An EMS integration project, which combined the Ireland very fast reserves called Fast Frequency Response, etc.)
and Northern Ireland EMSs into one new system, has will require online monitoring to alert grid controllers to
recently been completed. This allows all-island power inadequate service provision in real-time and to instigate
system operation from either control centre, and provides remedial action.
Figure 13 A sample table showing estimated largest frequency nadir and zenith and maximum and minimum RoCoF together with relevant contingency name
These outputs will be fed into the Unit Commitment Tool 8. Bibliography
to produce the final dispatches that have all operational [1] EirGrid. Operational Constraints Update 21st November 2016.
security criteria met. [Online]. http://www.eirgridgroup.com/site-files/library/EirGrid/
OperationalConstraintsUpdateVersion1_46_November_2016.pdf
6.2. Voltage Management [2] I. M. Dudurych, A. Rogers, R. Aherne, L. Wang, F. Howell and X.
Voltage management is becoming more challenging due Lin., “Safety in numbers,” IEEE Power & Energy Magazine, pp. 62-
70, March-April 2012.
to a vastly increased number of dispersed reactive devices,
each with different capability characteristics; and due [3] EirGrid, All Island TSO Facilitation of Renewables Studies. Dublin:
EirGrid, 2010.
to increased variability in generation and demand. New
tools are required to manage the change, one of which [4] M Burke, R Lopez, C Martin, and M Gallagher, “Advanced EMS
Wind Dispatch Tool for the Power System of Ireland and Northern
is a voltage optimisation tool to facilitate evolution from Ireland,” in Proc. 13th Wind Integration Workshop, 2014.
the current heuristic methods of voltage control. Such a
[5] R. Aherne, J. Conroy, D. Connolly, R. Doyle, I. Dudurych, H. Jones
tool will determine optimal reactive targets for different and A. Rogers,”Design and Implementation of a Tool for Assessment
types of devices, delivering voltage plans secured against of Secure level of Wind on the Irish Power System,” in Proc. 2010
contingency events for both day-ahead within day re- CIGRE Session, paper C2-107.
optimisation. More efficient methods of voltage control [6] EirGrid Group. Delivering a Secure, Sustainable Electricity System
will also be required. Whereas a few large synchronous (DS3). [Online]. www.eirgridgroup.com
Abstract one of the key challenges that we will have to face in the
coming years. A sustainable development approach must
GE’s Grid Solutions, through its Clean Grid program, has be applied, by responding to “the needs of the present
made a commitment to reduce emission of Green House without compromising the ability of future generations to
Gases (GHG) from its products. The target of this paper is meet their own needs”, as defined in the BRUNDTLAND
to present the two main axes of this program. report in 1987.
The first axis is the optimization of existing technologies For more than ten years, GE’s Grid Solutions has embarked
by reducing the impact of SF6. The paper presents the on the path of eco-design – integrating the environment
benefits of this eco-design approach carried out on the into the design phase alongside other conventional design
most recent generation of GIS. concerns such as customers’ expectations, technical
The second axis was to search for an SF6 alternative. performances, cost control, etc. Eco-design is a concrete
GE’s Grid Solutions, has developed an SF6-free solution solution to sustainable development. Indeed, curative
named g3 (green gas for grid), which is based on 3M approaches have already shown their limits paving the way
NovecTM 4710 and CO2. g3 reduces the global warming for an integrated, preventive approach. The eco-design
potential by 98% compared to SF6 with only minor design approach begins at the time of the definition of product
modifications from the equivalent SF6 design. The paper characteristics, because the design phase determines 80%
presents the different steps of the development, from of future impacts and offers the greatest degree of liberty.
the beginning in 2010 until the implementation in a first Through the deployment of its eco-design approach,
product: the 420kV Gas-Insulated Line. the objective of Grid Solutions is to reduce the global
environmental impact of the GIS, over its whole life cycle,
1. Introduction from the extraction of the raw materials until the end-of-
In the IEO (International Energy Outlook) 2015 life, while considering all environmental indicators.
projections [1], energy demand grows by nearly one-third
SF6 is, as a matter of course, one of these environmental
between 2013 and 2040.
aspects, having a global warming potential about
At the same time, the world is waking up to the reality of 23500 times greater than CO2. Life-Cycle Assessment
climate change. The Paris Agreement [2], signed by 195 performed on a gas-insulated substation (GIS) shows
nations worldwide at the COP21 conference in December that SF6 emissions that can occur during the service life
2015, brings all nations into a common consensus on of the switchgear can represent between 60% and 80%
their current and future responsibilities regarding the use of the total Global Warming impact of the substation
of carbon and addressing climate change. Together with over its whole life cycle. For the 17 others indicators the
interoperability and flexibility, reducing the environmental main environmental aspects are materials used and power
impact of our electrical transmissions systems will be losses during the use phase.
*elodie.laruelle@ge.com
KEYWORDS
Alternative Gases, Eco-design, g3,Gas Insulated Line, SF6
Having identified the importance of SF6 emissions, Grid This chapter presents the benefits of this approach when
Solutions, through the Clean Grid program, has made a applied to the latest generation of GIS regarding SF6
commitment to reduce these emissions from its products. impact: sealing improvement, reduction of SF6 quantity,
This paper presents the two main axes of this program: the reduction of seal lengths, optimization of assembly in
reduction of SF6 impact from the latest generation of SF6 factory, introduction of gas monitoring.
filled GIS, and the search for an SF6 alternative.
2.1. Sealing improvement
2. Reduction of SF6 impact from
The choice of the gasket and in particular its material
the latest generation of GIS is fundamental to ensure sufficient tightness at each
The first axis is the optimization of existing technologies interface, to ensure the reliability of the GIS during all
by reduction of SF6 impact from the latest generation its lifetime and not to exceed the maximum leakage rate
of GIS. While GIS users have a greater responsibility of 0.5% per compartment per year as specified in IEC
for reducing SF6 emissions during the use, maintenance 62271-203 [6].
and end-of-life phases, GIS manufacturers have the Early designs of Grid Solutions’ GIS in the 1960’s used
heavy responsibility of limiting the use and emissions simple O-ring type gaskets sealed with grease. In the
of SF6 during the design, as well as manufacturing and late 1970’s this technology was improved with a second
installation phases. sealing system added to improve tightness and limit
This is why reducing SF6 impact has been directly corrosion. This solution has proven to be efficient over
integrated in the eco-design approach developed in the long-term and has been implemented on many GIS
the R&D department [3]-[4]. Indeed, two of the six substations. It is still used in specific cases like the sealing
key environmental performance indicators (KEPI) are of accessories (filling valves, rupture disks etc.).
dedicated to SF6 impact: SF6 mass and seal length (see After more research and testing, an innovative type of seal
Fig.2). These KEPI have been created in order to set up for flange-to-flange connection emerged in the 1980’s
targets to reduce environmental impacts and assess our which has a particular shape delivering “three-seals-in-
progress, as defined in the “eco-design conscious process” one”. Two outer lip seals protect the huge inner sealing
described in IEC 62430 [5]. surface from atmospheric corrosion and from the effects
of SF6 decomposition products. The inner sealing surface
is well protected over the lifetime of the GIS.
This LCA shows that the use of g3 in the 420kV GIL allows Thus, National Grid wanted to undertake a pilot project
to considerably reduce environmental impact compared with the 420 kV GIL for its new Sellindge substation in
to the SF6 technology: Kent. The new Sellindge substation will be an integral
• on Global Warming, with a huge reduction of 96% part of National Grid’s transmission network and is an
• on all other environmental indicators (ozone important network reinforcement project. It is located in
depletion, acidification, metal depletion..), with an the South of England adjacent to the existing 2,000 MW
average reduction of 14% on all indicators. HVDC interconnector which links the British and French
electricity transmission systems.
3.3. Focus on the first pilot project in UK
The new substation consists of five feeders (four overhead
420 kV GIL will be the first g3 product to be energized. lines and one cable feeder to the interconnector). Two of
Thus, this chapter focusses on the first pilot project that these feeders, including 300 metres of gas-insulated busbar
is delivered and installed during first semester 2016 to on which g3 will be piloted, are separable from the rest of
National Grid utility in the UK. the substation using a bus section. In the unlikely event
National Grid has a strong commitment to managing all of a problem with the pilot installation this arrangement
of its environmental impact and actively promotes and provides a level of risk mitigation which is important when
supports initiatives, such as g3, to identify ways to reduce introducing new technologies onto operational transmission
SF6 usage across its network. Furthermore, under the RIIO systems. The fact that the g3 hardware can be relatively
(Regulation = Incentive + Innovation + Outputs) regime easily converted back to operation with SF6 also contributes
in the UK continuous reduction of SF6 losses is financially to the overall risk assessment undertaken by National Grid
incentivised. Taken together these drive a strong interest prior to embarking upon the live pilot at Sellindge.
in being at the forefront of adopting commercially viable
methods of reducing or eliminating SF6.
Assessment of harmonic disturbance in offshore wind power offshore projects is expected to be even longer, perhaps
plants (WPPs) is becoming an increasingly important task beyond 150 km. Experience shows that operation of
as the WPPs are increasing in size. It has been customary such a long cable system can have a significant impact
in the past to base all compliance and design studies on on power quality in terms of harmonic background
positive sequence simulation models. However, the use amplification due to resonances caused by the offshore
of long high voltage (HV) cable connection systems WPP electrical infrastructure as well as the onshore
gives rise to the need for more sophisticated modelling. and offshore transmission system. During low power
It is justified that in case of unsymmetrical cable systems, production or in some cases under normal production,
such as flat formation single phase HV cable systems a the background harmonic distortion in the onshore grid
decoupled sequence model can lead to underestimation of can potentially be amplified towards the offshore WPP
the harmonic distortion in the system. leading to potential design or compliance issues both at
The electrical connection to the 400 MW Anholt offshore the PoC and internally in the park array cable system.
WPP is used as a study case example. Synchronized Such harmonics amplification at different locations
harmonic voltage measurements were taken at the Point within the WPP and transmission system electrical
of Common Coupling (PCC) for Anholt (400 kV) and infrastructure can possibly bring concerns in relation to
at the Point of Connection (PoC) (33 kV). Based on the estimated harmonic voltage distortion levels or limits
measurements, corresponding per phase harmonic voltage specified in grid codes.
gains are determined. Furthermore, the harmonic voltage The importance of the phenomenon is gaining momentum
sequence components are determined at the PCC of Anholt. due to the growing number of large-scale WPP connection
It is shown that at and near resonances the harmonics do into the transmission grid. In Denmark, the HV export
not follow their natural sequence order, but the voltages system is owned and operated by the transmission system
contain some portion of all three components (i.e. positive, operator (TSO) and therefore harmonic compliance is
negative and zero sequence components). evaluated and specified at the PoC designated as on the
secondary side of the park transformers at the offshore
1. Introduction platform. Hence it is the responsibility of Energinet.dk
An AC cable system comprising of submarine and land as the Danish TSO to ensure power quality at the PoC is
cable sections is the most common way of integrating a compliant before the connection of the WPP. However, due
large offshore WPP to the onshore transmission network. to the combined length of the array cables and other passive
The length of the combined cable system can, in some components within the WPP electrical infrastructure,
cases, be very long – in Denmark for instance the amplification can increase due to the shift in resonance
longest WPP cable connection is 84 km. Furthermore, conditions or introduction of new resonances when the
the array cable system interconnecting the individual WPP is switched in. Therefore it is equally important
wind turbines (WTs) to a collection point can be over for both the TSO and the WPP developer to understand
100 km long making the total length of the cable system and limit the cause of the background harmonic voltage
close to 200 km. The total length of HV cables for future amplification due to resonances within the electrical
* cfj@energinet.dk
KEYWORDS
Cable crossbonding, Cable modelling, Electromagnetic compatibility, Harmonics, Harmonic amplification, Harmonic
propagation, Inter-Sequence coupling, Phase domain,
harmonic voltages can be v unbalanced. An unbalanced With both the harmonic magnitudes and phase angles
harmonic order will contain a portion of all tree sequence correctly determined for each phase, the sequence
components rendering the former assumption invalid. components of the harmonic voltages can be determined
A typical Danish transmission level harmonic spectrum using Fortescue transform. This is done for the harmonic
from the 2nd to the 20th harmonic taken at Trige 400 kV is voltages shown in Figure 2 and the results are presented
shown in Figure 2. The signals used for the analysis are in Figure 3.
measured using the specially developed high bandwidth The sequence decomposition of the harmonics shows that
high voltage sensors described in Section 2.1 there is a tendency that most of the harmonics follow their
Figure 2 shows strong unbalanced behaviour for some of natural sequences under the balanced system assumption.
the harmonics and more balanced behaviour at others. For For instance the 2nd order harmonic is predominantly a
instance, the magnitude of the 11th order harmonic voltage negative sequence harmonic, the 4th is predominantly a
on phase B is almost twice as large as the magnitude of positive sequence harmonic, the 5th is predominantly a
phase C. Also, the 3rd order harmonic, often assumed to be negative sequence harmonic and so on. However, the 11th
a pure zero-sequence harmonic, shows strong unbalanced order harmonic, that is expected to be negative sequence,
behaviour indicating a diverse content of the three contains a non-negligible positive sequence component.
sequences. The 3rd order harmonic contains all three sequence
components with a slightly higher content of positive several harmonic frequencies. Even at power frequency
sequence. Hence, it is safe to conclude that the assumption the impedance of the middle laid phase can be 12-15%
that individual harmonic orders contain only one unique lower than the two outer phases [4]. Near resonance, the
sequence component will lead to an underestimation of phase impedance of one cable core conductor can be
the highest phase-to-ground harmonic voltage which can significantly lower or higher than the two others, causing
be an issue for planning and design purposes. The reason unsymmetrical harmonic currents to flow in the system
for the unbalanced behaviour at some harmonic orders is even under balanced harmonic injection. As a consequence
analysed in Section 4. of the unbalanced series-impedance matrix of the flat
formation cable, pure positive sequence injection can,
4. Phase domain harmonic besides energizing the positive sequence system, also
energize the negative sequence system and the zero
modelling sequence system to some degree, with unsymmetrical
Harmonic propagation studies are most often conducted phase voltages as a consequence.
before system components are selected because changes
A full detail frequency dependent phase-domain model of
to the system are either not an option or very expensive
the Anholt grid-connection cable system and array cable
after the installation process begins. Hence, the designer
system is constructed in DIgSILENT Power Factory. All
relies on trustworthy simulation results for design- and cables are modelled based on a geometrical description
compliance studies. Such harmonic design studies are and all cross-bondings are manually implemented. The
often carried out in the decoupled sequence domain where frequency dependence of the transformers and reactors is
positive sequence and sometimes zero-sequence data is included using appropriate multiplication factors [5]. The
used to build a simplified harmonic model of the system WTs are modelled using a Thévenin equivalent including
under study. Such modelling approach can be inaccurate turbine transformers and WT internal filters.
on a system with a significant amount of cables laid in
flat formation due to the inherent unbalance introduced To illustrate the importance of phase domain modelling,
naturally as indicated by the measurements in Section an ideal positive sequence harmonic voltage is applied
3. Furthermore the detailed phase-domain modelling behind the system impedance at Trige 400 kV at each
requires appropriate representation of harmonic sources integer frequency from the 2nd to the 20th harmonic
(e.g. harmonic background distortion measured at Trige and the three phase-to-ground voltages and the positive,
400kV) including information about phase amplitudes negative and zero-sequence voltages at the 220 kV
and angles. offshore substation are calculated as shown in Figure 4.
A phase-domain model is necessary to predict the correct The resulting phase impedance as seen from Trige 400
behaviour of the harmonic voltages in an unbalanced kV (Zp(h)=Up(h)/Ip(h)) where p is equal to A, B or C ) is
system like the Anholt WPP. The land cable part of the presented in Figure 5.
electrical connection is comprised of three single-core By comparing Figure 4 and Figure 5, it can be clearly
cables cross-bonded and laid in flat formation. Such a flat seen that the application of positive sequence voltage
formation cable system introduces strong asymmetry at only predicts well the harmonic phase-to-ground voltage
at frequencies far from resonance. This is at conditions The harmonic voltage gains of the three sequence
where the phase impedances of the three phases are components are well defined due to the definitions
balanced as seen from the location of harmonic injection. governing the decoupled sequence domain. There is a
At resonance and near resonance (i.e. harmonic range linear relationship between the sequence components of
from the 3th to 4th and between the 9th and 14th in Figure 5), the voltage at the two busses; increasing the harmonic
the three phase impedances differ significantly and hence distortion at Bus A by some factor will lead to an increase
inter-sequence coupling will occur. Notice for instance the at Bus B by the same factor in a one-source system. This
negative sequence voltage at the 12th harmonic in Figure is one of the main benefits of the decoupled sequence
4 is larger than the positive sequence voltage – this is domain when it comes to harmonic propagation studies.
under positive sequence injection only into the system as The same cannot be said in the phase domain. The per
explained. Notice also that at the 13th harmonic, the phase phase harmonic voltage gain would naturally be defined
B voltage is more than double in magnitude compared to as:
the magnitude of phase A voltage. In sequence domain
this manifests itself as a much lower (almost half) positive
sequence voltage implying that assessment based on
positive sequence will lead to underestimation of the
harmonic voltage. In a decoupled sequence networks where UpX(h) and UpY(h) are the harmonic voltage of
based model this phenomena would be missed altogether. phase p (A, B, or C) at the hth harmonic at Bus X and Y.
Therefore, using only the positive sequence component Many TSOs use a decoupled sequence model for
for compliance analyses at planning/design stage can be harmonic studies. However, harmonic emission limits
problematic in unbalanced systems. are issued and measured in relation to the highest phase-
to-ground value. Because mutual coupling between
5. Phase domain harmonic the three phases differs in an unsymmetrical conductor
voltage gain configuration (such as that introduced by a cable
system laid in flat formation), the decoupled sequence
During harmonic compliance studies it is common assumption is violated. In the decoupled sequence
practice to calculate harmonic voltage gains to establish domain, the three sequences are independent of each
effect from one node to another. In the decoupled other. However, the harmonic phase voltage gain depends
sequence domain, the positive sequence harmonic on the unique flows in the other system conductors. This
voltage gain from Bus X to Bus Y at the hth harmonic makes the calculation of the phase harmonic voltage
can be determined as: gain complicated in an unbalanced system.
To illustrate the error introduced by using per default
the decoupled sequence domain and hence positive
sequence gains, the harmonic phase and positive
where U1X(h) and U1Y(h) are the positive sequence sequence voltage gains from Trige 400 kV to the reactor
harmonic voltage at the hth harmonic at Bus X and Y station GNK220 and the gains from the reactor station
respectively. The negative and zero sequence harmonic to the off-shore platform are determined (see Figure 1
voltage gains are defined in the same manner but not for system diagram and bus references). The results are
often used in practical studies. shown in Figure 6.
Figure 6(a) shows that the per phase and the positive 13th order harmonic voltage phase gains from Trige 400
sequence harmonic voltage gains from TRI400 to GNK kV to Anholt 33 kV are calculated based on measurements
220 (asymmetrical flat cable formation) strongly differ in over two days. The 13th order harmonic voltage as function
magnitude. This is more pronounced at and near resonance of time (10 min avg. values per IEC 61000-4-30) are shown
frequencies. in Figure 7(a) and the resulting harmonic voltage gains
calculated using Eq. 2 in Figure 7(b).
The displayed phase and positive sequence gains in
Figure 6(b) confirm that on symmetrical conductor Figure 7(b) shows that the measured 13th order harmonic
configurations (such as the submarine cable laid for the voltage gains differ by a more than a factor of 2 between
Anholt WPP), the positive sequence gain can be used the three phases. Unfortunately, due to the inherit
without loss of information. Notice the small deviations limitation of the MV IVTs, the phase voltage angles cannot
for instance at the 19th and 20th harmonic in Figure 6(b). be determined at Anholt 33 kV and hence the positive
sequence gain cannot be calculated and compared to the
The deviations are caused by the 0.5 km flat formation
phase gains. However, with the displayed differences
beach cable used between the reactor station and the
between the lowest and highest phase gains, the positive
beach cable/submarine cable joint.
sequence gain will not correctly represent all three phases
To emphasise the point at an individual harmonic level, the as discussed in the previous section.
Figure 7 (a) Trige 400 kV and Anholt 33 kV harmonic 13th order voltages and (b)
corresponding harmonic voltage gains.
The positive sequence harmonic voltage gain in the as more and more cables are installed at transmission level
decoupled sequence domain is per definition not dependent and the degree of (harmonic) unbalance thereby increases.
on the negative or zero-sequence component and vice The MV array cables and most often the HV submarine
versa. However, the harmonic voltage phase gain can be cables are symmetrical due to construction characteristics
strongly dependent on which sequence components are and hence the phenomenon of inter-sequence coupling is
energized if the system is unbalanced. This is illustrated in less important. However, even with a relative short land
Figure 8 where two cable systems are energized by a 1 pu cable section laid in an unsymmetrical formation in the
balanced harmonic voltage at all harmonics from the 2nd to system, inter-sequence coupling should not be ignored.
the 20th. Gradually, a harmonic phase voltage unbalance is
introduced by increasing the negative sequence content of 6. Conclusions
the voltage from 0 pu to 1 pu while the harmonic positive the paper discussed the modelling approach for harmonic
sequence voltage component is kept constant. One cable propagation and assessment studies on systems consisting
system is laid in touching trefoil formation (symmetrical) of long EHV cable systems. The 400 MW Anholt offshore
and the other in a flat formation (unsymmetrical). The WPP is used as a case study example.
harmonic voltage gain from Trige 400 kV to Anholt 33
kV for phase A is then plotted for both cases. Decomposition of high accuracy measurements taken at
the PCC of Anholt WPP into sequence components shows
Inter-sequence coupling is absent on the cable system in that individual harmonic orders in general follow their
touching trefoil system as the harmonic phase impedance natural sequences (that is negative for the 2nd, zero for
matrix is balanced (in contrast to flat-formation system) the 3rd, positive for the 4th and so on). However, this is
and as a result, the harmonic sequence impedance matrix not the case at and close to resonances as each individual
is dominated by the diagonal elements (off-diagonal harmonic will contain some portion of the two other
elements are close to zero). Therefore, the per phase sequence components. This is expected to be the case in
harmonic voltage gain is not dependent on the magnitude almost all systems due to the inherent unbalanced nature
of negative sequence harmonic voltage applied. For the of the power networks. However, the situation may be
unsymmetrical cable system, the per phase harmonic exacerbated in certain situations and especially when
voltage gain is strongly dependent on the magnitude of there is a high degree of natural unbalance introduced
the negative sequence voltage applied (or present) at and around the assessment location. The analysis has shown
near resonance due to inter-sequence coupling as seen in that when modelling highly unsymmetrical cable systems
Figure 8(b). Additionally, there will be dependence on the as is the case with EHV and HV cable systems laid in
angle of the sequence components, complicating the issue flat formation, determination of voltage gains to a remote
further, but this is not covered here. node or amplification of background distortion at the local
node can vary depending on the complexity of modelling
Phase-domain modelling becomes increasingly relevant assumed.
*hanspeter.hoschle@energyville.be
KEYWORDS
Capacity mechanism, Market adequacy, State of affairs, Survey results,
System adequacy, CIGRE working group C5.17
Figure 2: System size of contributions (based on data from 2012/2013)(Abbreviation see Table III)
_______________________
1- This information was collected in 2014 with respect to capacity mechanisms in place (CM (present)) as well as planned or discussed capacity mechanisms (CM
(planned i.e. up to 6 to 10 years ahead) which can either be plans for the introduction or the revise of an existing mechanism. In the following, present will always
refer to a period 2012/2013 for observed data or 2014 for present market systems in place, and planned refers to “from 2014 on”.
annual consumption (in TWh) given for the years as part of their market design. Strategic reserves or
2012 respectively 2013. Both winter ( ) and summer operating reserves are implemented in 7 markets
( ) peak systems are represented. A direct link to the (AU NEM, FI, PL, SE, US PJM, US NYISO)2. They
existence of a capacity mechanism cannot be identified. are intended to cope with seasonal extremes (“cold
Systems with and without capacity mechanisms exist reserve”) or as “safety net” if demand cannot be met
both with summer or winter peak. Also the size of a at the market. In case of US PJM and US NYISO, the
system cannot be seen as a direct indicator for or against reserves are part of a combination with other market
the need of a capacity mechanism. based capacity mechanisms. Capacity payments are in
place in 5 markets (ES, GR, IE, IT, CL). In AU WA,
Similarly, the ratios between day/night and winter/
CL, US NYISO and US PJM capacity payments are also
summer are compared (see technical brochure for
part of a combined capacity mechanism. US ISONE has
numeric values [1]). The ratios represent a rough
implemented a capacity auction. Capacity auctions are
estimate of the daily respectively seasonal demand
also part of the market design in US NYISO and US
volatility. These indicators are valid for categorizing the
PJM. A capacity mechanism based on reliability options
power system but similar to the peak demand and annual
is implemented in Colombia. A visual overview is given
consumption, a direct link cannot be drawn. However, for
in the map in Figure 3.
the winter/summer ratio, the bigger the deviation from 1,
the more seasonal a system is, which might lead to less Planned Implementations
operating hours of peaking units and an (emerging) need
for a capacity mechanism (France, Finland, Sweden). In 10 of the 31 jurisdictions discussion is ongoing about
On the other side, this is not the case for example for the redesign, introduction or abolition of a mechanism.
Norway. Therefore, the extended assessment in the BE, FR and GB are in the process of introducing a
technical brochure of the working group also includes capacity mechanism coming from a market design
the share of installed capacities respectively generated based on an energy-only market. BE introduces strategic
energy per fuel type. This is in particular interesting for reserves to cope with seasonal shortages in the winter.
the evaluation of systems that are dominated by one GB and FR chose for a market-based approach, namely
resource, i.e. nuclear-, fossil fuel-, and hydro- or RES- capacity auction respectively ex-post capacity obligation.
based systems. Moreover, no direct link can be derived Discussions in DE and JP have not yet reached consensus
from system specification in terms of generation mix about a possible introduction, not to speak of a type of
and the existence of capacity mechanism. For example, mechanism. CL, IE and IT are in the process of changing
hydro-based systems exist both with (Brazil, Colombia) their existing capacity mechanism design. In all the 3
and without capacity mechanism (Norway). jurisdictions the transition is made towards a reliability
options design. SE has fixed a time line to phase-out the
3. Capacity mechanisms strategic reserves until 2020 again. This is illustrated in
Figure 4.
worldwide
4. System Reliability
Present Implementations
In all answers, the responsibility for the long-term
The evaluation of present capacity mechanisms is generation adequacy and short-term security of supply is
limited to the jurisdictions with a history of capacity borne by the government, the system operator or a panel
mechanisms for several years. In total, 16 out of 31 _______________________
market areas have implemented a capacity mechanism 2- Abbreviations for jurisdictions are listed in Table III in the appendix.
Country AU AU BE BR CL CN CO CZ DE ES FI FR GB GCCIA GR IE
NEM WA
System Reliability
Present ++ ++ + - - ++ ++ ++ ++ ++ ++ ++ ++ - ++ ++
3 years ahead ++ ++ - - - ++ - ++ ++ + +/- ++ ++ - ++
Capacity Mechanisms
Present ()
Planned/Discussed
None
Main reason
Other •/- - - - - - - - - - - - - • -
Country IL IN IS IT JP NL NO PL SAPP SE US US US US US
CAISO ERCOT ISONE NYISO PJM
System Reliability
Present ++ + + + +/- ++ + ++ - ++ + + ++ ++ ++
3 years ahead ++ ++ ++ + ++ ++ - - ++ - - + ++ ++
Capacity Mechanisms
Present
Planned/Discussed
None
Main reason
Network - - • - - - • - - - - - - -
Other - - - - - - - - - - - - - -
that have at least one of the two parties as member. This is considered. At the same time, next to the generation
allocation clearly shows that the long-term responsibility adequacy assessment, the sizing of interconnections and
also in liberalized market design is not only left to the internal network planning is part of the security of supply
outcome of the market decisions. The system operator studies.
has hereby the role of an approved expert and counsellor
The received answers for system reliability is given below
on future capacity demand expectations while the final
responsibility and authority to set rules is with the in Table I. The table shows the information received for
government. In most cases, the system operators conduct present and expected system reliability together with the
a certain monitoring and assessment. However, the existence of capacity mechanisms. In all jurisdictions
answers indicate that there are hardly procedures that are with concerns about the system reliability, the concern
found common in all jurisdictions. Common predefined is connected with the generation adequacy. Only in three
measures that are repeatedly used are for example the jurisdictions the network adequacy causes concerns about
Loss of Load Expectation (LOLE), accompanied by e.g. the system reliability. On the long-term perspective, the
simulations of severe weather conditions such as dry years concerns increase with the unpredictability of the system
or cold spells. Mostly, such simulations and calculations development.
lead to annual or multi-annual adequacy / security of Jurisdictions with a capacity mechanism in place
supply reports that form baseline for policy making. tend to have a positive expectation about their system
If applicable for the jurisdictions, interconnections are reliability in the future. Jurisdictions that are discussing
taken into account for the assessment of security of or implementing a capacity mechanism at the moment
supply. Especially in case of system simulations and have been woken up by concerns about the generation
scenario calculations, the impact of the interconnections adequacy in the near future.
Country AU AU BE BR CL CN CO CZ DE ES FI FR GB GCCIA GR IE
NEM WA
Liberalized
Power pool
Bilateral contracts
Day-ahead markets
Real time/Balanc-
ing
Ancillary services
Capacity
Mechanisms ()
Country IL IN IS IT JP NL NO PL SAPP SE US US US US US
CAISO ERCOT ISONE NYISO PJM
Liberalized
Power pool
Bilateral contracts
Day-ahead markets
Real time/Balanc-
ing
Ancillary services
Capacity
Mechanisms
Asia
Dhabi, Oman)
country-specific constellations of the described criteria.
27 IN India
References 28 JP Japan
29 SAPP South African Power Pool
[1] G. Doorman, J. Barquin, L. Barroso, C. Batlle, A. Cruickshank, C.
rica
Af-
Dervieux, R. Flanagan, J. Gilmore,
30 AU NEM Australia National Electricity Market
J. Greenhalg, H. Höschle, P. Mastropietro, A. Keech, M. Krupa, J.
Austra-
Riesz, B. LaRose, S. Schwenen, 31 AU WA Western Australia
lia
G. Thorpe, K. De Vos, L. de Vries, and J. Wright, Capacity
mechanisms: needs, solutions and state of affairs. Technical Brochure
647, Paris, CIGRE, 2016. Acknowledgement
Appendix Thanks to all country experts for providing answers
in the worldwide survey. Special thanks to the CIGRE
Table III: List of submissions
C5.17 working group members for feedback on the
# Code Jurisdiction survey design and evaluation of the answers.
1 BR Brazil Biographies
America
South
2 CL Chile
Hanspeter Höschle holds a Diplom degree in
3 CO Colombia
industrial engineering from the Karlsruhe Institute of
4 US CAISO California Independent System Operator Technology, Germany. Since 2012, he has worked as a
5 US ERCOT Electric Reliability Council of Texas Ph.D. researcher at the KU Leuven, Belgium, on power
America
6 US ISONE ISO New England Reliability system economics and the modeling of electricity
North
Coordinator
markets, in particular capacity mechanisms. He holds
7 US NYISO New York ISO
a Ph.D. fellowship of the Research Foundation
8 US PJM PJM Interconnection
- Flanders (FWO) and the Flemish Institute for
9 CZ Czech Republic
Technological Research (VITO). He was an active
10 BE Belgium
member in the CIGRE C5.17 working group and
11 ES Spain
organized the worldwide survey. He already discussed
12 FI Finland
various aspects of capacity mechanisms both at the
13 FR France
CIGRE symposium 2015 and the Paris Session in 2016.
14 DE Germany
15 GB Great Britain Gerard Doorman has a in Electric Power Engineering
16 GR Greece
from the Norwegian University of Science and
Europe
* gkorres@cs.ntua.gr
KEYWORDS
Terms - Adaptive protection, directional overcurrent relays, distributed generation, distribution systems, hardware-in-
the-loop simulation.
and failed reclosing phenomena are discussed in detail. which frequently follows an out-of-zone fault, or due to
Section 3 presents and describes the architecture of the DG backfeed to a fault on an adjacent feeder [1], [2]. This
proposed adaptive protection infrastructure, introducing phenomenon is also called false tripping, and belongs to
also the concept of optimal setting group determination. the wide class of nuisance tripping problems due to various
Proof of concept and results obtained by applying the root causes (e.g. motor starting, magnetising inrush current
proposed approach to a distribution test system with large during transformer energisation, cold-load pickup, voltage
DG penetration are analysed in Section 4. Finally, Section sag or swell etc.). In this paper, we focus on the sympathetic
5 highlights the conclusions of this work. tripping scenario caused by the contribution of DG for
faults beyond its feeder protection zone.
2. Impact of distributed generation As has been discussed in recent literature, DG units can
on feeder protection cause false tripping and undesirable disconnection of an
adjacent healthy feeder. The basic principle of sympathetic
2.1. Protection Blinding tripping can be explained in Fig. 2, where a fault occurs
at Feeder 2 and the DG unit connected to Feeder 1 feeds
Traditional distribution systems had only one source
the short-circuit through the substation bus. If the DG
contributing to a fault incident, that is the main substation,
contribution is particularly large, relay R1 may operate
and protective relays were set to see a certain distance down
before relay R2 takes action and clears the fault. This
the radial feeder. This distance is usually referred to as the
is possible when non-directional overcurrent relays are
reach of the protective device, which is actually determined
used for feeder protection, which cannot discriminate
by the minimum fault current that the device can detect.
the change of fault current direction (reverse fault). The
However, the connection of DG units to modern distribution
utilisation of non-directional overcurrent relays is a
networks introduces additional fault current sources, which
long-established protection practice for most distribution
may increase the total short-circuit level within the network,
utilities owing to their single-fed radial networks, unlike
while altering the magnitude and direction of fault currents
modern multi-source distribution systems with bi-
sensed by installed protective devices.
directional current flow.
When the DG plant is located between the utility substation
and the fault location, the total fault current is increased
due to the partial contribution of DG, as illustrated in
Fig. 1. On the other hand, the fault current seen by feeder
relay R1 is actually decreased for the same fault, owing to
fault current division between the sources, which may not
exceed the pickup current setting of R1. This undesirable
protection performance is widely known as protection
blinding [1], [2]. The blinding effect results in delayed
protection operation or even total desensitisation in case
of weak upstream system and large DG penetration. This
phenomenon is also called protection under-reach, since
the actual reach of the feeder relay is decreased due to DG
fault current contribution, as also shown in Fig.1. Fig. 2. DG contribution to fault incident on adjacent feeder
(sympathetic tripping case)
case of DG presence. A reclosing attempt is deemed to be temporarily. The breaker re-closes its poles after a
successful when there is sufficient time margin between sufficient time delay, to allow transient faults to extinguish
shots for the fault arc to dissipate and clear, which means and re-establish cold non-ionised air at the fault point. In
that each DG unit must detect the presence of a fault and case of a permanent fault, the recloser should operate with
disconnect early in the reclose interval, as illustrated in its slow curve, using a time-delayed overcurrent element
Fig. 3. Otherwise, the DG unit will continue to feed the 51P/N and allowing the load-side fuse to clear the fault.
fault hindering the arc extinction, and the fault that would Fuse-saving schemes are based on the premise that both
have been temporary becomes permanent. recloser and downstream fuse sense the same current.
A potential reclosing failure to clear the fault means that However, the DG presence in distribution networks
some customers will suffer a sustained outage, whereas increases the total fault current, and thus the branch fuse
they should be subjected to only momentary interruptions. will always see more current than the upstream recloser
Thus, the reliability of the power delivery system is slightly (Fig. 4). As a result, the fast curve will be shifted to the
degraded. Furthermore, the active power unbalance right on the time-current characteristic (TCC) diagram,
during the dead time of the reclosing sequence can lead to and the fuse may blow before the fast operation of the
frequency variation in the islanded part of the distribution recloser, as depicted in Fig. 5.
grid, and generators may drift away from the synchronism
with respect to the main system [1], [2]. In this scenario,
a reclosing attempt would couple two asynchronously
operating systems. Moreover, conventional reclosers are
designed to reconnect the circuit only if the substation
side is energised and the opposite side is de-energised. In
case of DG integration, there would be active sources on
both sides of the recloser, hampering proper operation of
the reclosers.
In addition to the aforementioned protection challenges,
the constantly increasing presence of DG units in Fig. 4. Fuse-saving scheme in DG-penetrated distribution system.
(DNOs). As a result, the entire performance of the systems was verified. The main feeders were protected
distribution system protection can be greatly enhanced. only by the non-directional relays R3 and R7, according
Various approaches have been proposed in recent to the traditional protection practice. The obtained HIL
literature [8]–[10]. The NLP approach was considered test results revealed the inadequacy of existing protection
for the suggested system and solved by state-of-the-art settings, which were determined considering single-fed
optimisation packages. An extensive presentation of the radial configuration for the distribution system.
problem formulation and description of the employed
More specifically, when the simulated network was
optimisation tools can be found in [11].
operating without DG presence and a three-phase fault
4. Test results was conducted at Bus 1.2., relay R3 operated properly after
0.554 s, seeing the total fault current (2.12 kA) supplied
To evaluate the performance of the proposed adaptive solely by the upstream external grid. Then, the same short-
protection scheme, a 5-bus distribution network with circuit test was conducted with DG1 connected. The total
variable operating mode was simulated, representing a fault current was increased due to DG presence, however
simplified two-feeder portion of the Hellenic Distribution the partial contribution from the external grid was reduced
System (Fig. 7). The phase overcurrent elements (51P) to 932 A, resulting in a greatly delayed tripping time of
of feeder relays were enabled and tested during the relay R3 (2.23 s). In addition, a symmetrical fault was
evaluation procedure, operating with the IEC C1 standard applied at Bus 2.1, causing faster operation of R3 than R7.
inverse characteristic. Detailed system and protection Stated another way, R3 was tripped sympathetically to R7
data are given in [11]. for an out-of-zone fault, disrupting the service continuity
The evaluation procedure was actually composed of three of the healthy line L1. The previous description is made
stages. In the first stage, the adaptive logic was inactive, clear in Fig. 8, where the TCCs of R3 and R7 relays are
while the likelihood of protection blinding and sympathetic plotted for each test case, showing also the sensed fault
tripping occurrence in DG-penetrated distribution currents and the corresponding tripping times.
(a) (b)
Fig. 8. Protection blinding (a) and sympathetic tripping (b) of R3 shown on the TCC diagram.
Table I. Simulation results for the 5-bus test system with DG1 connected (SG1 active)
Table II. Simulation results for the 5-bus test system with DG1 & DG2 connected (SG2 active)
2
ABB AB, Corporate Research, Sweden
3
Institute of Electrical Power Systems and High Voltage Engineering, TU Dresden, Germany
4
ABB Switzerland, High Voltage Products, Switzerland
1. Introduction
Interfaces are critical locations in gas-insulated systems:
Failure occurs in the gas, but is often mediated by charges
accumulated on insulators: surface flashovers generally
follow a short path of high electric field. In practice,
this often means that the field component normal to
the insulating surface is critical for surface charge
accumulation, but flashover is mainly triggered by a high Figure 1(a), Components of a HVDC gas-insulated system: (1) conductor, (2)
enclosure, (3) insulating gas, (4) solid insulator, (5) capacitive screen, (6) insert,
tangential field component [1]. The temporal evolution of (7) current collector, (8) dielectric coating on metallic surfaces. Electrical effects
the DC electric field in the system is mainly determined for the design of a HVDC insulator in gas-solid insulation: electronic and/or
ionic conduction, polarization, space charge accumulation. In gas insulation: Ion
by conduction in the solid insulation. The radial potential generation by (a-d) and ion drift (e); On interfaces/surfaces (f-j). (b), Design op-
distribution changes with time from capacitive to timization scheme for HVDC insulators. In red: optimization constraints on the
electrical volume conductivity σ of the insulator material. Other constraints are in
resistive-controlled. This causes a field redistribution, as blue, and potential modification of components in orange color. En and Et: electrical
the location of highest electric field on the insulator-gas field components normal, respectively tangential to the epoxy insulator-gas surface.
*robin.gremaud@ch.abb.com
KEYWORDS
Charge emission, epoxy insulation, gas insulated system, high voltage direct current (HVDC), surface potential.
[10] and D the diffusion coefficient. The diffusion of the domain, and otherwise. The development
coefficient is related to the mobility via Einstein’s relation of surface charge on the solid-gas interface (“insulator
the Boltzmann constant, T0 = 300 K is the surface”) is thus determined by the continuity equation:
gas temperature.
A direct solution of these equations is numerically (9)
challenging for complex geometries, so we instead use
a model inspired from Ref. [11]. In this model, only the where is the current in the solid. Electric field
ionic current from the gas impinging the insulator surfaces and surface potential ϕ are then calculated
is solved for, and not the ion concentration in the whole using the Poisson equation:
gas volume. In practice, this is done by solving the steady- (10)
state drift-diffusion equations(6) together with the source
term for natural ionization (7): Whether gas or solid charging current dominates on the
interface depends locally on the ratio of normal currents.
constant (7) If natural ionization is the only source of gas ions, this
for each ion species in the gaseous domain at every time ratio is where L is the
step, making use of the fact that the time needed for newly length of the field line ending on the insulation surface
created ions to drift to a surface is much shorter than all and En the normal field component in the solid [2, 11].
the other relevant timescales in our problem. Therefore, the solid tends to dictate the field distribution
for high voltages, unless the insulator geometry creates
large capture volume or charge injection and multiplication
2.3 Charging of gas-solid interface, surface potential
dominates. This ratio determines also the sign of the
Gas ions located within the capture volume accumulate surface charge and of the remaining surface voltage when
on the insulator surface where they build a surface charge. the main voltage is switched off.
Similarly, space charge from the solid insulation also
tends to accumulate on the other side of the interface. 3. Experimental measurments of
Accumulated charges typically reduce the normal
component of the field at the insulator interface at the side electrical currents in insulation
from which they reach the interface. Eventually, they can 3.1 Solid insulation: DC conduction
also contribute to conduction in the solid.
Figure 2 shows current curves for polarization-
The charging of the gas-solid interface is modeled by depolarization experiments of alumina-filled epoxy
introducing the following boundary current source on the composite cast into an aluminum electrode system,
insulator interface: including guard-ring. This cast block geometry offers
(8) several advantage respectively to the previously used
contacting system of conductive rubber on epoxy plates
where ng is the normal vector pointing from the surface [2]: namely i) the same metal/insulation contacting as
of the insulator in the gas and the current in the real insulator application ensures realistic charge
density from the gas phase. injection behavior at contact, ii) reduced quantity of void
So-called “open” boundary conditions are applied to the at contact compared lead to lower partial discharge noise,
other interfaces (electrodes and grounded tank), i.e. using and ii) less influence of short-time temperature fluctuation
in regions where the charges flow out (increased thermal mass).
As seen in Figure 2, in this case an estimation (within Small electric fields were applied (< 2 kV/mm) in order
20% error) of the DC current is to detect the onset electric fields of charge injection and
already achieved after one day (8.6·104 s) of measurement multiplication.
time, while a steady state necessitates 11 days (106s) of
Figure 3 shows a simplified sketch of the measurement
measurement. By repeated polarization/depolarization
system: High voltage (positive and negative polarity, up
measurement at all operating temperatures, the Arrhenius
to 100 kV) is applied to the “injecting” electrode while
behavior of the DC conductivity
another electrode (collector) collects these charges. The
is determined. The conductivity value varies over three
collector is connected to a current measurement system.
decades from 20°C to the maximum operating temperature
The size of the electrodes, i.e. the volume in between was
of 105°C [4]. As a consequence, an inhomogeneous
chosen in order to reach approximately 1-2 pA due to
temperature distribution results in space charges in the
natural ionization.
solid and field redistribution under DC compared to AC
stress. (11)
where V is the volume of the gas. With a value of =
3.2 Gas insulation
37.5 IP/(s·cm3) [13], the current due to natural ionization
Although the physical mechanisms of charge injection in the given volume (~ 0.168 m3) is ~2 pA. Although the
and multiplication are well known, the absolute number natural ionization current is dependent on the volume, it is
of charges and the onset-fields can vary in orders of plotted as current density (A/m2) to represent an “injected
magnitude. The biggest unknown in the system – which charge”. The surface of the inner conductor is ~1 m2,
can also not be described analytically – is the electrode therefore the absolute values of current and current density
surface roughness. Therefore the amount of injected approximately coincide. The onset for charge injection
charges for a technical relevant geometry was investigated can be located in a range between 0.3 and 0.7 kV/mm.
experimentally (see Ref. [12] for a description of the This value is much lower than values from literature [14,
experimental setup). In order to separate gas dominated 15]. Emission currents at higher electric fields are shown
effects from electrode-dominated effects the experiments in the right panel of Figure 4, in both SF6 and vacuum, for
were conducted using both vacuum and pressurized SF6. negative polarity.
Figure 4 Left: Emission current at low electric fields, and right, at higher fields, for negative polarity and smooth (rms roughness Ra = 3±0.2 μm) emitting electrode
both for vacuum and 0.45 MPa SF6.
4. Surface potential measurement The time-scale of the transition to the resistive state
(which depends mainly on the conductivity of the solid
Surface potential measurement are performed with insulation) is well-predicted by the model5. For early
a modified non-contact electrostatic voltmeter from times (t < 300 h), the measured potential increases in
TREK, Inc, using a miniature electrostatic field chopper most of the cases more rapidly than predicted, which is
probe [16]. A positioning system for the probe allows to be expected as fast polarization currents are neglected.
scanning of two insulator surfaces within a 320 kV The amplitude of the change in surface potential ∆U is
DC gas-insulated system (see Figure 5). Tested are DC however controlled by charges in the gas: more charge
insulators under 0.45 MPa SF6 at relative humidity RH injection/multiplication reduce the final potential change,
≤ 8%. Measurements are performed at full load current, as can be seen in Figure 6c). In this particular case,
resulting in an average 25°C thermal gradient across the considering natural ionization only, the expected increase
insulation3. To speed up the transition to DC state, the in surface potential would be ∆U=13 kV on the insulator
minimum enclosure temperature is lifted up to 50°C. surface, corresponding to a steady-state injection current
After system thermalization, a voltage UDC = 100 kV is
= 0.35 pA/m2 into the insulator surface. In contrast, the
applied on the central conductor and the surface potential
maximum experimentally observed ∆U(ϑ=−30°)= 8 kV,
of the gas-insulator interface is measured periodically at
corresponding to an additional injection current
various azimuthal angles ϑ and radii r until steady state is
10 pA/m2 from the metallic enclosure, corresponding to
reached.
= 2.2 pA/m2 onto the insulator surface.
Figure 6a) shows the temporal evolution of the measured
These relatively low additional ionic currents reduce the
and modeled surface potential change at an average radius
change in radial distribution between the capacitive and
r = 120 mm. For modeling of the potential evolution,
the steady state DC electric field (see Figure 6c).
DC conductivity values of the alumina-filled epoxy
insulator material as function of temperature and field are
determined independently by polarization experiments on
5. Flashover breakdown voltage
plates and cast blocks4. The only adjusted parameter in After proper experimental calibration of the DC simulation
the model is the injection/multiplication current (see model at nominal field values, its applicability should be
Figure 6b)). Effective values used are within the tested at near-breakdown conditions. For this purpose,
same range as measured in section 3.2 and Ref. [14] for a setup for the determination of the flashover voltage
comparable surface roughness (Ra, see Ref. [17]). of cylindrical epoxy specimens subjected to a thermal
___________________________
3- Due to natural convection, the thermal distribution depend on the azimuthal angle. A separate measurement of the temperature distribution as function of load current
is performed prior to voltage switch-on by a network of thermocouples distributed on the insulator surface.
4- A direct scale-up of the fast polarization current obtained from thin samples (plates/cast blocks) to this real-size component is difficult and thus neglected here.
5 An exception is the upward direction (0°). In this location the natural convection is considerably higher than anywhere else, and the associated gas velocity field might
interfere with the field-driven ion drift.
gradient at 0.1 MPa SF6 pressure was built and operated scaling the flashover voltages Vbd,DC after long-term stress
at TU Dresden [18]. with VDC for (6, 48, 92) hours.
Figure 7 a) Equilibrium temperature distribution b) Calculation of capacitive-resistive transition after DC energizing at voltage VDC with
temperature distribution neglecting charge multiplication. (calculation 1 in Figure 8)
Under warm conditions, all long-term tests show a show a moderate reduction of the minimum flashover
decrease of the breakdown voltage in comparison to the breakdown strength of 0.8 · Vbd,VRT,warm. This behavior
values of the voltage rising test and show the effects of the can be explained with calculation 2 by the presence of
capacitive-resistive transition. After 48h, the insulation low injection ionic currents from the neighboring HV
strength is on average about 10 % lower than in the VRT. electrode. The magnitude of these currents is comparable
Due to the temperature-dependent conductivity of the to those reported in Ref. [14].
insulator material, the highest field strength is increasing
and its location is shifted to the cooler part of the insulator 6. Conclusion/outlook
into the gap near the high voltage electrode (Figure 7b). This contribution presents a joint experimental/theoretical
approach for the determination of electric potential at the
5.3 Comparison with simulation gas-solid interface in HVDC gas-insulated systems. The
associated model enables prediction of the spatial and
The simulation of the capacitive-resistive transition
temporal evolution of the electric field.
reproduces the execution of the experimental test (see
Figure 8, right): Using i) Curie-von Schweidler approach The input experimental parameters needed for
to simulate the polarization processes, and ii) the modeling are the transient and steady-state current-
independently measured DC conductivity in the epoxy voltage characteristics of the insulation media, i.e. the
resin, the electric field is calculated for a DC voltage temperature-, field- and time-dependent residual DC
application of various pre-stress durations. Subsequently, currents in gas and solid insulation. These are determined
a fast voltage rise is implemented in the FEM tool, by independent measurements on model experimental
until reaching the breakdown field strength on the HV setups: for the solid, epoxy plates or cast blocks, for the
electrode. Considering conduction in the gas, simulation gas, a coaxial electrode geometry.
was carried out for two cases: The time scale of the capacitive-resistive surface potential
1. Calculation 1: residual homogenous conductivity of transition is controlled by the DC conductivity of the
10-21 S/m in the whole gas volume. This approximation solid epoxy insulation. The timescale predicted by the
is justified, as the contribution from natural ionization independently-calibrated model is in good agreement with
is negligible for this voltage level and geometry. the measured temporal evolution of the surface potential.
This result is essential to derive reliable testing procedure
2. Calculation 2: as calc. 1, with additional field-dependent of DC gas-insulated system. Furthermore, the applicability
injection current = of the model for the prediction of breakdown voltage as a
10 pA/m2) on the insulator interface subjected to high function of DC stress duration is demonstrated.
field next to the HV electrode.
Further investigations should aim at completing the
Comparing the calculated values with the measurements characterization of charge injection and multiplication
(Figure 8a)), agreement is found for both cases for times in the gas, with e.g. the effects of paint, various
of DC stress up to 100 h. For longer times, calculation 1 surface roughness, and gas humidity content. Effects
predicts that flashover should happen after twelve days on conduction in the solid, such as long-term thermal
under pure DC stress at VDC. However, two experiment and electrical aging, as well as the scalability of “fast”
have been carried out at VDC for 14 days, respectively polarization currents to real-scale insulation components
27 days without breakdown. Subsequent VRT applied deserve also further study.
KEYWORDS
Load profiles; Dynamic clustering; Pattern recognition; Classification.
730000 load profiles being analyzed). Figure 1 depicts energy consumption per day, ranging from a maximum
the patterns obtained after applying the dynamic peak of around 0.3 kWh in households from cluster 3,
clustering algorithm previously described. The number to a maximum peak of energy consumption of around
of clusters to be found is a parameter that must be 2 kWh in households from cluster 4. The seasonality
defined. A number of 6 clusters to be found has been effect can also be observed in the six clusters obtained,
chosen, based on previous experiences with the data. with an elevated rate of energy consumption in Winter
(around January and February months).
Observing the rasulting patterns in Figure 1, some
quick conclusions can be derived. It can be observed, Time series clustering allows, therefore, obtaining
for instance, that there are four clusters which gather the patterns that evolve through time, in a time frame
majority of households (clusters 1 to 4), and there are defined by the expert. This allows an interpretation of
two clusters with probably unusual energy consumption the results that depicts the full dynamic behavior of all
patterns and a more reduced number of customers or the objects, allowing a much more complete (and also
households belonging to them. These are clusters 5 and complex) global view of the resulting temporal patterns.
6, with 54 and 11 customers respectively. Moreover, Following, the resulting patterns are characterized
observing clusters 1 to 4, another conclusion could by means of specific quantitative indices that allow
also be obtained: that the shapes of the four clusters are to describe the resulting patterns in their evolution in
very similar, and their main difference is in the average shape and through time.
Figure 2. Graphical representation of the quantitative indices defined to describe the temporal patterns obtained,
in terms of shape analysis (a) and the evolution of the load profile through the days (b). Cluster 1.
Table II. Results of classification of new customers and classification errors per index.
Nikoleta Kandalepa, Dr. Jasper van Casteren, Dr. Susana Almeida de Graaff*, Danny Klaar
TenneT TSO B.V.
Arnhem, The Netherlands
KEYWORDS
Flow-Based method, network investments evaluation, network sensitivities, optimization, zone-to-zone transmission
capacity
2. Flow-Based Method calculation process. The IGMs of the TSOs are merged into
one network file, which contains the common grid model
The FB method is applied two days before the execution (CGM) with the best forecast for one hour of the execution
day to calculate the day-ahead zone-to-zone transmission day. Based on a DC load flow computation, the FB method
capacities. The resulting FB domain represents the computes the power transfer distribution factors (PTDFs)
maximum available capacity for market transactions in the and the remaining available margin (RAM) for the CBCO
day ahead market [4]. set. These constitute the so-called FB parameters.
Every day, each participating TSO creates its individual Each RAM equals the remaining flow capacity of a single
network model (IGM) based on the best available forecast CBCO. The zone-to-zone PTDFs (as defined in section
for the execution day regarding vertical load, grid topology, 3.1.3) equal the change in RAM caused by a variation in
renewable infeed and available production. Two days ahead zone-to-zone power exchange for each combination of two
of the execution day, these models are provided to the FB specific zones.
computation, together with files containing the generation
The RAM of a CBCO is the maximum capacity of the
shift keys (GSKs) and the combinations of critical branches
branch (PmaxCB) minus the reference flow of the CBCO
(CBs) and critical outages (COs). The Generation Shift (FrefCBCO):
Key (GSK) defines how a change in the net position (NP)
of a bidding zone is mapped to the generating units of that
zone [5]. Equation (1) describes the RAM as used in the FB for
Critical Branches (CBs) are a subset of all network investments concept.
elements that are considered significantly impacted by During day-ahead capacity calculations, non-costly
zone-to-zone exchanges. These branches are taken into remedial actions (RA) may be applied in a coordinated
account in the capacity calculation and, consequently, in way, such as changing the tap position of phase shifting
the market allocation process. The CBCO set is built upon transformers (PST) or topological actions.
the critical branches, taking into account both N and N-1
situation by adding the combinations of a critical branch Figure 2 shows an example of a FB domain that is defined
by the FB parameters. The FB domain determines the
with a critical outage (CO). Therefore a single CBCO may
market constraining CBCOs and the maximum allowed
be a single CB or the combination of a CB with a CO,
zone-to-zone exchange. The FB domain is impacted by
when this combination is considered relevant to the zone-
both changes in RAM and PTDF of each CBCO and
to-zone market capacity. The CBCO set thus represents
these are normally both altered by investment proposals.
the monitored network constraints from a zone-to-zone
Therefore, the evaluation of an investment proposal needs
capacity perspective.
always to address both the impact on RAMs and zone-to-
Figure 1 shows a simplified representation of the flow-based zone PTDFs for all CBCOs.
The nodal PTDF can be aggregated into zonal PTDF by where PTDFX,tCO and PTDFX,fCO are the nodal PTDFs for
adding the effect of the MW power shifts of the individual the “to” and “from”-nodes of the CO, and “X” represents
generators in one zone. The “Generator Shift Key” (GSK) the CB or the CO.
is the participation factor of a single generator in the MW The LODF is the change in MW power flow of a CB due
change of the net position of a whole zone. A zone is to the outage of another branch. The LODFCB,CO is used to
therefore defined by the set of GSKz for which calculate the power flow of the CBCO from the
=1, and the zonal PTDFCB,z is calculated as: pre-outage power flows of the critical branch ( ) and
the critical outage ( ):
The flow-based market coupling in the CWE region has
four specific zones: Belgium, Germany, Netherlands and
The LODF matrix has a row for each critical branch and a
France, but any definition of zones or GSK can be used.
column for each critical outage.
The zonal PTDF of a CB shows how a change in the net
3.1.5 PSDFs
position of this zone influences the power flow of
this CB (PCB): The impact of the phase shifting transformers is linearized
by using the phase shifter distribution factors (PSDFs).
The PSDF expresses the change in CB flow for a change
Any change in the net position of only one zone (in radians) of the tap angle of a PST:
is balanced by an opposite change at the slack node. A
different slack node will thus produce different zonal
PTDFs.
where i is the PST index.
Zonal PTDFs for all zones are computed for every CB.
The PSDFs are directly calculated from the nodal PTDF
For the four CWE countries in the FB market coupling,
matrix through:
four zonal PTDFs per CB are thus computed.
3.1.3 Zone-to-zone PTDFs
The PSDF matrix has a row for each branch in the network
By subtracting the zonal PTDFs from 2 specific zones, and a column for each selected phase shifting transformer.
the corresponding zone-to-zone PTDFs are calculated. The PST tap positions are normally discrete, and the
The zone-to-zone PTDF of a CB (PTDFCB,z2z) shows how PSDF describes the impact of any tap position.
a change in the bilateral exchange between these zones
influences the power flow of this CB: For DC load flow calculations, the PSDFs are mutually
independent and the PTDFs are independent from the
PSDFs. This is the basic assumption for linear PST tap
A balanced exchange between two zones does optimization that is described in section 3.4.
where the first line is the objective function that states that
Equation (20) directly relates the PST tap positions to α is to be maximized and the change in tap positions (τ)
the zone-to-zone transmission capacity. Any constraining is to be minimized with Cτ as penalty. The second line
CBCOs can be directly related to the PSTs and the states that the new RABEM must be higher than α for
4. Results
A dataset of 816 hourly network scenarios from 34 days Figure 7: Constraining CBCOs from B to A, non-opt. PSTs
was analysed. Each scenario contains the European The same results, but now including the optimization of
network model, and the capacity between four zones (A, PST tap positions, are shown in figure 8. In comparison
B, C and D) is analysed. These models are operational with figure 7, the RABEM is higher for both base and
congestion forecast network models. investment case. The effect of the PST optimization is to
A new tie-line between zone A and zone B was added to more evenly utilize the individual assets. Normally the
each base case model, to create the investment case for PSTs are used to redistribute the asset loading throughout
every hour. the network. In flow-based for investments, the effect
of the PST optimization is to balance the zone-to-zone
The RABEM and ARBEC indicators were computed for capacity margin, per CBCO, resulting in a higher capacity
each timestamp, for every zone-to-zone direction, for that can be made available to the market.
both base and investment case, and with and without PST
optimization. Figure 8 shows the same CBCO (nr. 16) to be the most
constraining for both base and investment case.
The calculation of RABEM and ARBEC provides
transparency and insights into the impact of the
investment on the zone-to-zone transmission capacity.
Reports containing results for single timestamps as well as
aggregated reports for large datasets have been evaluated.
Figure 7 presents the RABEM results for the 15 CBCOs
with the lowest RABEM values for the direction of
zone B to zone A, for one timestamp, for both base and
investment case, without PST optimization.
The CBCOs were sorted to the lowest RABEM for either
base or investment case. The most constraining CBCOs
Figure 8: Constraining CBCOs B => A, opt. PSTs in B => A
for both cases become visible in that way. In order to
evaluate the added value of the investment, the smallest The observed significant impact on the capacity margin
RABEM in the base case (left arrow) is compared with by PST optimization, means that the controllability of the
the one in the investment case (right arrow). The smallest PSTs is significant for both cases for this specific optimized
RABEM is higher in the investment case, and the new tie- zone-to-zone direction. Moreover, the optimization of
line therefore increases the transmission capacity margin PSTs may change the constraining CBCOs. The most
for this zone-to-zone direction, for this timestamp. The constraining CBCOs in figure 8 (after optimization) are
in the direction from zone B to zone A, separately for The flow-based method has the advantage of combining
the base and the investment case. The dots move to the the different zone-to-zone transmission capacity evaluation
right compared to figure 13, as the base case RABEM is into one analysis. Additional exchange between zones
increased by the optimization. However some timestamps A and B may reduce the remaining capacity between
show little impact of the PST optimization and remain other zones. Figure 15 combines the change in RABEM
close to zero. The majority of dots also move upwards. for two different zone-to-zone directions and shows the
The Δ(RABEM) scatter plots of figure 13 and 14 produce interdependencies between these two directions. The
a clear view on the overall performance. In both figures, horizontal axis represents the Δ(RABEM) from zone B to
the majority of the dots are located in the upper area, A. The vertical axis shows the Δ(RABEM) from zone B
to C. No optimization of PST is applied.
which means that the investment has added value for most
timestamps. Figure 15 shows a upper right quadrant for the timestamps
for which the investment results in a higher available
capacity margin for both zone-to-zone directions. The
lower left quadrant shows the timestamps where the
transmission capacity is lower for both zone-to-zone
directions. The other quadrants show timestamps where
the transmission capacity is lower for one direction and
higher for the other.
Conclusions
A new concept for the evaluation of investment
alternatives has been developed, which can be used in
different planning timeframes: short-, medium- and long-
term planning. This concept is applicable to real network
Figure 14 : BC minRABEM vs Δ(RABEM), B => A, models for the whole interconnected European grid, and
opt. PSTs from B => A offers a transparent insight into the impact of investments
By plotting the scatter plots for all relevant zone-to-zone on the transmission network. The concept also bridges the
directions, and by adding quantitative measures such as gap between network planning methods and the market
average displacement, extreme values, cluster mid points, capacity calculation method.
etc., they can provide a clear insight into the benefits The proposed capacity performance indicators, RABEM and
and risks of an investment proposal, when the amount of ARBEC, quantify the impact of a single network component
analysed scenarios is sufficiently large. on the overall remaining zone-to-zone transmission capacity.
Any type of investment (e.g. new line, new tie-line, new [8] J. Guo, et al., Direct Calculation of Line Outage Distribution Factors,
IEEE Transactions on Power Systems, Vol. 24, No. 3, August 2009
PST, new generator) can be evaluated. The analysis of the
[9] P. Hoffmann, et. al., The simple tie-line decomposition method – a
impacts of the phase shifting transformers is included in new approach for a causation based cost-sharing key, Cigre Science
the concept. Linear optimization of the PST tap settings & Engineering, N°5 June 2016