Cigre - 544 - Metal Oxide (MO) Surge Arresters - Stresses and Test Procedures
Cigre - 544 - Metal Oxide (MO) Surge Arresters - Stresses and Test Procedures
Cigre - 544 - Metal Oxide (MO) Surge Arresters - Stresses and Test Procedures
MO Surge Arresters
Stresses and Test Procedures
Working Group
A3.17
August 2013
MO SURGE ARRESTERS
WG A3.17
Members
B. Richter, Convenor (CH),
J.L. De Franco (BR), R. Ghler (DE), F. Greuter (CH), V. Hinrichsen (DE), M. Holzer (AU), S. Ishibe
(JP), Y. Ishizaki (JP), B. Johnnerfelt (SE), M. Kobayashi (JP), K. Lahti (FI), T.M. Ohnstad (NO), R.S.
Perkins (CN), M. Reinhard (DE), J.H. Sawada (CA), A. Sironi (IT)
Corresponding Members
A. Dellallibera (BR), R. Diaz (AR), S. Vizintin (SL), Y. K. Tong (GB)
Copyright 2013
Ownership of a CIGRE publication, whether in paper form or on electronic support only infers right
of use for personal purposes. Are prohibited, except if explicitly agreed by CIGRE, total or partial
reproduction of the publication for use other than personal and transfer to a third party; hence
circulation on any intranet or other company network is forbidden.
Disclaimer notice
CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept
any responsibility, as to the accuracy or exhaustiveness of the information. All implied warranties
and conditions are excluded to the maximum extent permitted by law.
ISBN: 978-2-85873-239-5
MO Surge Arresters
STRESSES AND TEST PROCEDURES
Table of Contents
EXECUTIVE SUMMARY ............................................................................................................. 4
Foreword ........................................................................................................................................................................... 8
1.
2.
Page 3
EXECUTIVE SUMMARY
The Cigr Technical Brochure TB 60 was published in 1991 describing effects on gapless metal oxide surge
arresters (MO arresters) from various electrical stresses encountered in 3-phase AC systems. Since then,
continued improvements in equipment technologies coupled with the interest of the de-regulated power industry to
maximize utilization of the existing infrastructure has revolutionized the MO arrester applications and the expected
performances in an environment, characterized by higher stress levels.
Todays proven confidence in the reliability and capability of modern MO arresters offers new possibilities of
overvoltage protection and improved management of power system disturbances.
The Working Group A3.17 of SC A3 took the task to evaluate the stresses on MO arresters and to review the
existing test procedures. Further on, the actual state of MO arrester designs was investigated, as well as the
various applications in different types of electrical networks.
Emphasis was given to the MO resistors as the active part of the MO arresters. A research project was started to
experimentally investigate the energy handling capability of the MO resistors, which is a key design criterion for a
reliable arrester application. The resulting Technical Brochure covers and describes the actual MO resistor and
arrester technology and the results of the first part of the research project on the energy handling capability of MO
resistors.
Electrical stresses on MO arresters can be divided into stresses at power frequency, which can have long time
durations, and transient stresses of short time duration resulting from switching and lightning. IEC 60071-4
proposes some recommendations for the evaluation of overvoltages, based on the use of numerical programs.
The different stress types seen by a MO arrester are:
Temporary Overvoltages
A temporary overvoltage (TOV) is an oscillatory phase-to-ground or phase-to-phase condition that is of relatively
long duration and is undamped or only weakly damped. TOV are one of the most crucial stresses to an MO
arrester and are detrimental for its layout.
The following origins of TOV are typically considered:
-
Earth fault temporary overvoltages occur in a large part dependent on the effectiveness of system earthing.
Guidance for the determination of TOV amplitudes is given in IEC 60099-5 and IEC 60071-2.
Disconnection of a load will cause the voltage to rise at the source side of the operating circuit breaker. The
amplitude of the overvoltage depends on the disconnected load and the short-circuit strength of the feeding
substation. The amplitude of load rejection overvoltages is usually not constant during their duration.
Accurate calculations have to consider many parameters.
Voltage rise along long unloaded lines (Ferranti effect).
Harmonic overvoltages, originating from e.g. DC converters or saturated transformers.
Resonances, in particular Ferro resonances.
Overvoltages due to flashover between two systems of different system voltages installed on the same
tower.
Slow-front overvoltages
Slow-front overvoltages, in most cases generated by switching or faults, are associated with load switching or fault
clearing. Different switching cases have to be considered: line re-energization, switching of capacitive loads and
inductive loads.
Page 4
Fast-front overvoltages
Fast-front overvoltages are in many cases caused by thunderstorms and occur all over the world. The heaviest
thunderstorms with the most intensive lightning will normally be experienced in the equator region. Other sources
are, for instance, current chopping of breakers or back flashovers.
In low voltage (LV) power systems up to 1 kV and medium voltage (MV) power systems (1 kV < Us 52 kV) the
distribution lines are generally of lower height and less exposed to direct flashes than transmission lines. Most of
the occurring overvoltages are then due to induced voltages originating from lightning to surrounding structures.
High voltage (HV) systems in the range of 52 kV < Us 245 kV can be found in transmission and sub-transmission
rural areas. Direct strokes, back flashovers and induced overvoltages will statistically result in a higher stress for
the installed arresters than in other voltage systems.
Transmission lines in extra-high voltage (EHV) with 245 kV < Us 800 kV and ultra-high voltage (UHV) systems
above 800 kV have steel towers with shield wires and are in spite of their height above ground well protected
against direct lightning strokes to the phase wires. Most of the lightning will hit the towers or the shield wires, and
only shielding failures and back flashovers will cause a critical surge in the phase wire.
In general, in 90% of all cases the lightning flashes are negative flashes from cloud to ground. However, some
countries, such as Norway or Japan, experience rather often thunderstorms during winter. Typical weather
conditions to create the winter thunderstorms are strong winds from the west, which transport warm air from the
ocean to the mountains of the mainland. The typical positive lightning flashes of winter thunderstorms transfer
higher charge than negative lightning flashes, which are typical for summer thunderstorms.
HVDC networks
Since the late 1970s overvoltage protection of HVDC converter stations has been based exclusively on MO
arresters. This is due to their superior protection characteristics and their reliable performance when connected in
parallel to the sensitive converters.
The continuous operating voltage stress for HVDC MO arresters differs from that of a normal a.c. arrester in that it
consists of not only the fundamental frequency voltage but also of components of direct voltage, fundamental
frequency voltage and harmonic voltages, and high frequency transients. These waveforms require other
dimensioning rules for the continuous operating voltage and some specific tests of the MO arresters, e.g. the
accelerated ageing procedure, as described in the emerging IEC 60099-9. Furthermore, polarity reversals might be
an issue.
Ambient stresses
Mechanical stresses like seismic loads strongly affect the structure and materials used for the design of the MO
arresters. Vibrations as well as static and dynamic loads have to be considered and appropriate test procedures
have been developed accordingly.
Ambient stresses can be very different in the different regions of the world. Very cold climates with ice and snow
loads have to be considered as well as climates of high temperature and high humidity. Observations of biological
growth on the surface of polymer insulation have been made in various places. Three types of organic growth have
been identified: Algae, Fungi and Lichen. Despite all the reports of biological growth on the insulation in some
areas of the world there are up to now no known failures of MO arresters caused by it. Animal impact may be
another issue in some countries of the world, e.g. Australia, where cockatoos would nibble on specific types of
polymeric material.
Page 5
MO Resistors
Steady progress has been made over the last decades in MO resistor technology, their application in overvoltage
surge protection devices and the understanding of the basic mechanisms of nonlinear conduction, energy handling
capability etc. A lot of new insights have been gained, new physical phenomena have been observed, improved
and more consistent models have been developed and much progress has been made in simulations related to
materials and components.
The nonlinear conduction mechanism of the material can be traced back to individual grain boundaries in the
ceramics, which show a typical value of the switching or breakdown voltage UB of approximately 3.2 V 3.4 V each.
Combining many grain boundaries in series and in parallel within an MO element allows to scale the voltage and
current characteristic of an MO resistor. For a sufficient large number of grain boundaries, the field strength E and
current density J then describes the material characteristic more generally.
Design of MO arresters
Different basic design principles are used for high voltage arresters and medium voltage arresters. In the high
voltage field mechanical requirements are much higher than in normal distribution applications. For this reason
porcelain housings are still used besides the growing number of hollow core insulators, so called tube designs,
and direct molded designs. For distribution arresters in medium voltage systems, porcelain housings have rapidly
disappeared and the direct molded design is used almost exclusively today.
Energy handling capability of MO resistors
The energy handling capability is a key property of MO arresters and has many different aspects, which are only
partly or not at all reflected in the actual standards. At least, though this list may be not complete, they have to be
divided into:
-
For the impulse energy handling capability single impulse stress, multiple impulse stress (without sufficient
cooling between the impulses), and repeated impulse stress (with sufficient cooling between the stresses) have to
be considered.
Thermal energy handling capability, on the other hand, can only be considered for complete arresters, as it is
mainly affected by the heat dissipation capability of the overall arrester design, besides the electrical properties of
the MO block.
For a deeper understanding of the energy handling capability of MO resistors and the relevant parameters, the
working group A3.17 initiated a research project to evaluate the energy handling capability under different impulse
stresses such as rectangular impulse currents, sine half waves, alternating currents and double exponential highcurrent impulses. More than 3000 specimens of commercially available MO resistors from seven well established
American, European and Japanese manufacturers were tested. Two basically different sizes of MO resistors were
considered, one for application in high voltage arresters (Size 1: 40...45 mm in height, 60 mm diameter) and
one for application in medium voltage arresters (Size 2: 30...40 mm in height, 40 mm diameter).
For the tests with impulse stresses, an extended failure criterion, beyond simple visible damages, was introduced
for the first time to differentiate the various failure modes and to quantify early changes in the electrical material
characteristics. The a.c. tests were performed up to mechanical failure.
It turned out that for the a.c. tests up to failure the statistical evaluation gives better information on very low failure
probabilities compared to the impulse stress tests (characterized by their mean failure probabilities).
Some of the most important conclusions from the research program, as discussed in more detail in the TB, are:
-
Energy handling capability has generally been improved over the last decade by the established
manufacturers.
Page 6
Outlook
The follow-up working group of A3.17 (A3.25: Metal oxide varistors and surge arresters for emerging system
conditions) is working on:
-
Further aspects of the energy handling capability such as durability (repeated impulses) or combined
stresses
UHV arresters
Consequences of increasing the field strength of MO resistors
Long term ageing of MO resistors
Consequences of the axial temperature distribution in an MO arrester
Page 7
Foreword
Cigr Technical Brochure 60 (TB 60) METAL OXIDE ARRESTERS IN AC SYSTEMS written by Working Group 06
of Study Committee 33, published April 1991, describes the severity with which system parameters affect arrester
performance and how system performance is affected by the arresters. The main intention was to give detailed
information on the application of the new type of surge arrester at this time. In addition the IEC standards for testing
and application developed in parallel. Many of the results of TB 60 were incorporated at this times in the new IEC
standards for metal oxide surge arresters (MO arresters).
TB 60 addresses, besides some basics about the characteristics of the MO material, the application in high voltage
3-phase transmission systems with 50/60 Hz and MO arresters with porcelain housings only.
Since the 1990s the application of MO arresters has increased in general and, due to the relatively simple and
robust mechanical design of the MO arresters compared to the conventional gapped arresters with SiC resistors,
new applications have become possible.
Continuous basic research on the MO material as well as the introduction of polymeric materials for the housings of
MO arresters for all system voltages has brought new and deeper knowledge and new application possibilities at
the same time. Today almost 100% of the medium voltage MO arresters have a polymeric housing, porcelain types
are not produced anymore on large scale. In the high voltage field more than 50% of all designs are of the
polymeric type, with increasing share. This development has brought new possibilities and as usual new questions.
The mechanical and pollution performance is of course different for polymeric and porcelain designs.
The MO material itself has been studied continuously over the years, which has brought better understanding of
the overall characteristics and better MO resistors with respect to electrical characteristics, homogeneity, long term
stability and energy withstand capability.
The number of manufacturers of MO resistors and arresters has increased, as well as the application of MO
arresters. Nowadays MO arresters are installed in a.c. and d.c. power systems with very different voltage levels,
from 660 V d.c. in traction systems up to 800 kV d.c. in HVDC systems, up to 1100 kV a.c. in UHV systems, and
they are used in substations, in cable systems, as line arresters etc., to give only some examples. In parallel, the
application of zinc oxide based MO varistors developed into a mass market for low voltage and electronic
applications, but this development is not described here.
The continuous development and the field experience with the MO arresters made it necessary to review the actual
state of the technology as well as the validity of the existing standards for testing MO resistors and arresters. An
example is for instance the classification of MO arresters in line discharge classes. The line discharge classes for
MO arresters are based on the energy that may be stored in transmission lines of different system voltages. This
classification works well as long as only 3-phase transmission systems up to 550 kV system voltage are being
considered. Various new applications in all electrical power systems, including UHV and HVDC, traction systems,
distribution systems etc. makes it necessary to reconsider the classification according to line discharge classes. For
this reason a critical review of the existing international standards was performed with emphasis on the energy
handling capability of MO resistors. To get a clearer and deeper understanding of energy related to MO resistors
and arresters the working group initialized a research program on energy handling withstand capability of MO
resistors. For the first time several thousand MO resistors for medium and high voltage application from many
different manufacturers were tested up to the limits and relations between the type of current impulse stress and
the failure mode of the MO resistors were evaluated.
Following the title Evaluation of stresses of Surge Arresters and appropriate test procedures and the scope of
working group A3.17 of Cigr SC A3, High Voltage Equipment, the TB is structured in the following sections:
Section 1. Stresses on Surge arresters
describes in general the different types of stresses on MO arresters, which may influence the performance of the
arresters. Naturally the performance of polymer housed and metal clad arresters is different in many aspects to the
performance of the classical designs with porcelain housings.
Page 8
- Subsection 1.2 (following a general introduction in 1.1) gives an overview about the stresses in 3-phase systems
with specific attention to temporary overvoltages and switching overvoltages. This is of special interest for system
studies.
- The subsection 1.3 addresses the special case of the very different voltage wave shapes and related stresses in
HVDC systems. The performance of the MO arresters under voltage stresses different from pure a.c. or pure d.c.
needs careful consideration.
- The specific conditions of d.c. and a.c. traction systems are dealt with in subsection 1.4.
- Stresses from lightning are discussed in subsection 1.5. Lightning parameters are given and the severe and
special cases of winter lightning is addressed. Results from studies and evaluations about the occurrence of
lightning stresses in different systems are given as examples.
- Subsection 1.6 deals with various stresses from the ambient. This can be divided into static and dynamic stresses
and the severe case of seismic stresses, which is especially important for larger equipment with mechanically
sensitive internal design like SF6 gas insulated (GIS) arresters. Further on, long term stresses with pollution and
humidity, as well as very low temperatures and temperature cycles, are of importance if polymeric insulation is
concerned. Biological growth is addressed in brief.
- Finally, subsection 1.7 deals with the electrical and mechanical stress under overload conditions.
Section 2. Functional parameters and design of MO Surge Arresters
deals with material and design aspects of MO resistors and arresters, respectively.
Surge arresters constitute an indispensable means for insulation coordination in electrical power supply systems.
A general definition states that a surge protective device is a device that is intended to limit transient overvoltages
and divert currents. Two different principles exist: voltage switching devices based on a spark gap (which are the
old gapped arresters with SiC resistors), and voltage clamping devices based on varistor technology. In the high
voltage community the todays devices are of the voltage clamping type and are called MO surge arresters, or
shortly arrester. A MO arrester has, simply speaking, to protect important and expensive electrical equipment
against damages resulting from overvoltages.
- Subsection 2.1 gives details about the voltage-current-characteristics of MO surge arresters, shows the current
and voltage wave forms as specified and standardized in international standards.
- Subsection 2.2 provides an overview of the material science of the MO material, the production process, and
leads from the micrometer scale of a single grain boundary up to the complete MO resistor and arrester. Possible
failure modes of the MO resistor and the long term performance of the material are addressed.
- In subsection 2.3 the different design principles of medium and high voltage arresters are shown.
- MO surge arresters with designs adapted to specific applications are dealt with in subsections 2.4, 2.5 and 2.6.
Section 3. Energy handling capability of MO surge arresters
deals with the need of a critical review of the existing standards and gives the details of the research project.
- In subsection 3.1 (summary) and 3.2 (introduction) the motivation of the performed research project on energy
handling capability is given and the general results are summarized.
- Subsection 3.3 explains the different aspects of energy handling capability for MO resistors and complete MO
surge arresters.
- In subsection 3.4 the state of knowledge and the initiated research project on energy handling capability of MO
resistors is described in detail.
- Subsection 3.5 finally gives a critical review of the many different aspects of the energy capability of MO surge
arresters in international standards.
Page 9
Section 4. Summary
summarizes the work of WG SC A3.17 and points out the influence on the actual standardization work in IEC TC
37.
APPENDIX 1
gives an overview about Cigr Technical Brochures related to MO surge arresters and their application.
Considering the actual development and discussion in the field of MO surge arresters some subjects have been
addressed more in detail than others. For instance the electrical stresses on MO surge arresters for application in
HVDC systems are of increasing interest because of the increasing numbers of HVDC lines. Further, MO surge
arresters (and high voltage equipment in general) for application in UHV systems require special attention with
regard to pollution and seismic stresses and possible test procedures. The development and the variety of possible
arrester designs made it necessary to go into the details of the actual designs available on the market. Questions
regarding the long term stability and the energy handling capability of the MO resistors can only be dealt with when
the material properties are given in detail.
The content of this TB was discussed and agreed by all members of the working group. The sections were written
by one or more authors in charge. Each section starts with a short introduction to the specific subject and ends with
a short conclusion. Thats why each single section can be read by itself without necessarily reading the other
sections.
Page 10
1.
1.1 Introduction
Surge arresters are widely applied in HV and MV a.c. and d.c. systems. They provide overvoltage protection from
the generator in the power plant up to the end-user, including protection of substations, overhead-lines, and cables.
They are installed in fixed installations and in traction systems. Due to the world wide applications under very
different and sometimes severe ambient conditions a variety of stresses occur.
According to the title of the working group Evaluation of stresses of Surge Arresters and appropriate test
procedures the working group collected all stresses that can occur.
The addressed stresses, like electrical stresses from the system, from lightning and from ambient are described in
the following chapters.
1.2.1 GENERAL
CIGRE WG 06 published Technical Brochure TB60 in 1991 which describes effects on gapless metal oxide surge
arresters (SA) from various electrical stresses encountered in AC systems. Since then, continued improvements in
equipment technologies coupled with de-regulated power industrys interest in maximizing utilization of existing
infrastructure has revolutionized SA applications and expected performances in a more stressful environment.
Confidence in the reliability and capabilities of modern SAs and power electronic based equipment offer improved
management of power system disturbances. On the other hand, such power systems which allow increasing
number of distributed generations tapping into existing transmission and distribution circuits or circuits of different
voltage classes sharing common towers increase operational complexity and higher occurrences of network stress.
Page 11
For load rejection of generator transformers, the temporary overvoltages may reach amplitudes up to 1.4 p.u. for
turbo generators and up to 1.5 p.u. for hydro generators. The duration is approximately 3 s.
TOVs from following causes may also require consideration depending on the nature of the network:
- Voltage rise along long unloaded lines (Ferranti effect).
- Harmonic overvoltages, e.g. DC converters or saturated transformers.
- Back feed through interconnected transformer windings, e.g. dual transformer station with common secondary
bus during fault clearing or single-phase switched three-phase transformer with an unbalanced secondary load.
Resonance: Linear resonance can be either series or parallel which could involve both large currents and/or
voltages. Ferro-resonance modes may be sub-harmonic or harmonic with the latter producing higher temporary
overvoltages. Temporary overvoltages from resonance should not form the basis for the surge arrester selection.
The use of a surge arrester to damp out resonance is not effective and unproven.
Combination of TOVs: Combinations of TOVs such as earth faults and load rejection may result in higher
temporary overvoltage values than those from a single event. When combinations are considered sufficiently
probable, overvoltages from each cause have to be compounded taking into account actual system configuration.
Severe TOV Cases: Occurrences of severe or repetitive TOVs are possible when different voltage circuits share
common towers and flashovers occur between high and low voltage conductors and possibly with multiple re-close
operations. While it is practical to provide limited SA protection for moderate TOVs of short duration, more severe
and/or sustained cases could cause multiple equipment and SA failures. To safeguard against such events it might
be worthwhile to install some designated SAs with lower protective level to relieve stress on parallel SAs and
equipment.
Page 12
considered necessary to ensure line switching performance and reliability when controllers are out-of-service or
CBs misoperate.
Page 13
cases, SAs may be applied directly in parallel with the switch to provide more effective TRV control but continuous
or temporary overvoltages expected during open switch conditions must be carefully examined.
Page 14
1.3.1 INTRODUCTION
The most significant differences for arrester applied in HVDC systems compared to normal AC-applications are the
wave shapes of the actual COV and TOV, and that the verification of the energy stresses becomes more complex.
For some applications the most severe energy stresses are sometimes not even followed by any significant service
voltage so that thermal instability cannot occur.
For the arresters indoors, for example in the valve hall, insulation withstand tests are not relevant and should be
skipped.
Continuous operating voltages
As the HVDC arresters can be applied at a variety of different positions, with many different wave shapes of the
service voltage, it is unpractical to give the voltage values in r.m.s. values. Therefore the continuous operating
voltage always has to be given in crest values together with the wave shape. Hence additional definitions for the
operating voltages are needed.
-
CCOV, which is the highest crest value of the continuous operating voltage excluding possible
commutation overshoots.
Non-significant CCOV, which is a continuous operating voltage of such low amplitude that the power losses
generated can never initiate a thermal run-away after energy injections. Each manufacturer will have to
give its limit for significant power losses of their designs.
PCOV, which is the highest crest value of the continuous operating voltage including possible commutation
overshoots.
ECOV, which is the equivalent ac- or dc-voltage, used in operating duty tests, having at least the same
power losses as the actual CCOV at the actual temperature after energy injections.
At the AC-yard there are applications apart from the normal AC-bus, e. g. in the filters. These HVDC-arresters may
rd
th
th
th
be stressed with sinusoidal voltages but with higher frequencies, like for example 3 , 5 , 11 , 13 harmonics of the
nominal power frequency.
In the converter station there are more complex wave shapes. Typically there are also commutation overshoots.
The wave shapes for valve arresters have a dc-component and a very short voltage peak of the opposite direction.
In the valve hall there may also be so called bridge arresters across 6- or 12-pulse groups. Their operating voltage
has commutation overshoots together with a high dc-component but with no voltage peak of opposite polarity. It is
important to notice that the commutation overshoots are more or less influenced by the arresters themselves
depending on the valve set up, so studies should be performed also with the arresters present to see the true
commutation overshoots, as the arresters may damp them significantly.
In the DC-yard there may also be filter arresters at different frequencies apart from the DC-bus arresters which see
a pure dc-voltage. Other applications may be across the smoothing reactor, which see a non-significant CCOV.
There may also be neutral bus arresters in the valve hall or in the DC-yard. They also have non-significant CCOV.
Accelerated ageing tests should be performed with wave shapes similar or obviously worse than the actual ones.
Also for determination of power losses at the actual wave shapes it is necessary to generate a variety of different
wave shapes in order to get the proper ECOV to use in operating duty tests. One way to solve this is to generate
similar wave shapes on the low voltage side and then amplify this voltage with an amplifier up to appropriate levels
for testing on individual ZnO-discs. For the filter applications a frequency generator can be used.
Worth noticing is also that some HVDC system may reverse the dc-polarity even after a long time with one polarity.
This means that there has to be shown in accelerated ageing tests, after 1000h testing, that the ZnO-discs, which
are exposed to CCOV with a high dc-component, can cope with this without causing excessive power losses or
other ageing effects.
Page 15
Energy stresses
The aim of the type testing for HVDC arresters is to verify both energy capability of the ZnO-discs themselves as
well as to verify their thermal stability after maximum energy stresses followed by CCOV. The operating duty tests
should in principal follow the IEC procedures, but standard test parameters for line discharge classes acc. to IEC
are typically not usable.
System studies are nearly always performed, resulting in energy stresses and typical transient wave shapes. For
HVDC applications it is proposed that the long-duration current impulse withstand test is substituted by a high
energy impulse withstand test with 6 impulses of the maximum energy requirement from the system studies,
separated by one minute apart. Then it has to be decided which test wave shape that best cover the actual energy
stresses; sinusoidal, half sine wave shapes or rectangular, which can typically be generated in arrester test labs
for operating duty tests. Guidance which wave shape to use can be found in Chapter 3, Energy Handling Issues.
The operating duty tests should be performed similar to the switching surge operating duty tests for line discharge
classes 4 and 5 in IEC 60099-4, with test samples preheated to 60C, but with the two long-duration current
impulses substituted with one energy impulse having at least the same energy as calculated in the system studies.
In some cases the maximum energy stress may also come from two consecutive impulses and then two energy
impulses one minute apart are used in the testing. After the energy injection, the test samples shall be exposed to
an a.c.- or d.c.-voltage voltage that generates the same or higher power losses as the actual wave shape. If there
are TOV stresses, the calculated energy from these stresses should be generated either by adding it to the energy
impulse or test the test samples with an equivalent ac- or dc-voltage generating the same or higher energy. TOV
stresses of the a.c.- or d.c.-bus arresters can of course be tested with the actual voltage and duration. It is also
recommended to use the same three test samples in both the long-duration current impulse withstand test and the
switching surge operating duty test.
It is not unusual that for the most severe fault scenario the converter station is closed down afterwards. In these
cases there is no need to verify thermal stability after this energy injection. But if there are other fault scenarios
followed by CCOV also this has to be verified. So for some arresters there may be one high energy value verified in
the high energy impulse withstand test and a lower energy verified in the operating duty tests.
In some cases like the neutral bus arrester application there is never any significant CCOV, so all the disc-tests can
be performed on open ZnO-disc sections. For this application, which often consists of several parallel columns,
there may also exists very rare fault scenarios with so extremely high energy requirements that it is more
economical to use the arrester as a sacrificial device with a failure of one or two columns. In this case special
consideration to the short-circuit tests may be necessary, in order to easily facilitate a restart of the converter
station.
Since the late 1970s, overvoltage protection of HVDC converter stations has been based exclusively on metaloxide surge arresters. This is due to their superior protection characteristics and their reliable performance when
connected in series or parallel with other arresters.
The basic principles when selecting the arrester arrangement are that:
-
Overvoltages generated on the a.c. side should be limited by arresters on the a.c. side
Overvoltages generated on the d.c. or earth electrode line should be limited by d.c. line arresters and
neutral bus arresters
For overvoltages within the HVDC converter station, critical components should be directly protected by
arresters connected close to the components, such as valve arresters
Information about selection, application and testing of HVDC surge arresters is given in the Application guide for
metal oxide surge arresters without gaps for HVDC converter stations prepared by CIGRE working group 33/14-05
and published in 1986. Further information is given in IEC 60071-5 Insulation Coordination Part 5: Procedures
for high voltage direct current (HVDC) converter stations. Parts of the CIGRE guideline are already included in IEC
60071-5
Page 16
Chapter 1: Scope
Chapter 2: Metal oxide arresters characteristics
Chapter 3: Arrester schemes and stresses on HVDC converter station arresters
Chapter 4: Studies for determination of arrester stresses
Chapter 5: ZnO arrester to limit temporary overvoltages
Chapter 6: Rules for determination of arrester capabilities and arrester test requirements
Chapter 7: Arrester testing
CCOV
U dim
(equation 1.1)
2 U v0
Udio
Udim
U v0
no-load phase-to-phase voltage on the valve side of converter transformer, r.m.s. value
Page 17
Figure 1.1: Typical waveforms of continuous operating voltages at various locations in the converter
station. All individual figures show the voltage over time.
1.3.2.2 Sources and types of overvoltages
Overvoltages on the a.c. side may originate from switching, faults, load rejection or lightning. Overvoltages on the
d.c. side may originate from either the a.c. system or the d.c. line or from in-station flashovers or other fault events.
Page 18
for the insulation on the a.c. side of the converter (a.c. equipment): the highest value of operating voltage
expressed as the r.m.s. voltage phase-to-phase (IEC 60815);the minimum recommended creepage
distances are defined in terms of mm per kV (phase-phase). Typically the range is between 16 mm to 31
mm/kV.
for the insulation on the d.c. side of the converter (d.c. equipment): the d.c. system voltage for the
insulation to earth, or a corresponding average value of the voltage across the insulation for insulations
between two energized parts.
The trend in the industry for several years has been to use larger specific creepage distances in HVDC
applications. For example, creepage distances as high as 60 mm/kV have been used in HVDC systems. However,
such an increase in the specific creepage distance did not eliminate the external flashovers.
The specific creepage distance of 60 mm/kV in a d.c. system corresponds to about 35 mm/kV in an a.c. system.
The use of composite housings for surge arresters has been successful also with smaller specific creepage
distances.
For an indoor clean environment, a minimum specific creepage distance of about 14 mm/kV has been widely used
and has not experienced any flashover.
For both d.c. and impulse voltages the positive polarity has lower withstand voltage than the negative polarity.
Page 19
The amplitude of the current for which the protective level is specified, which is referred to as the co-ordination
current, is usually selected differently for different types of current wave shapes and locations of the arresters.
These co-ordination currents are determined from detailed studies carried out during the final stages of the design.
The arresters used on the a.c. side are usually specified as for arresters in a normal a.c. system by their rated
voltage and maximum continuous operating voltage.
For the arresters on the d.c. side of a HVDC converter station, the rated voltage is not defined and continuous
operating voltage is defined differently because the voltage wave shape which continuously appears across the
arresters consists, in many cases, of superimposed direct, fundamental and harmonic components and, in some
cases, also commutation overshoots. The arresters are specified in terms of:
-
This means that the tests specified for these arresters shall be adjusted for the particular applications, different
from standard tests usually applicable for a.c. arresters.
The required energy capability of the arresters shall consider the applicable wave shapes as well as the
amplitudes, duration and the number of respective discharges.
Page 20
Event
Arresters
A
Current extinction
X
X
Page 21
earth fault between the converter transformer and the valve in the commutating group at highest potential;
clearing of an a.c. fault close to the HVDC converter station;
current extinction in only one commutating group (if applicable).
Depending on current rating, control system dynamics, inductance of the d.c. reactor, and the protection scheme,
the phase to earth fault will be dimensioning for the energy and current rating of the arresters.
The valve arresters can in general only be subject to fast-front and steep-fronted overvoltages at back-flashovers
and earth faults within the converter area. The most critical case for steep-front overvoltages is normally an earth
fault on the valve side of the converter transformer of the bridge with the highest d.c. potential.
Page 22
Page 23
1.4.1 GENERAL
The overvoltage protection of the electrical traction systems has an increasing importance nowadays. This is not
only by the railways which are supplied with a.c. voltage but also increasingly by the d.c. railways.
The long-distance railway system is electrified with 3 kV d.c. voltage on over 70 000 km rails (in year 2000, that
means about 38% of the total length of the rails of the electrical railways) and with 1,5 kV d.c. voltage on more than
20 000 km (about 11%). That means that about half of the world-wide railway length of the long-distance traffic is
operated with direct-current. The length of the electrified rails by the outer suburban service, including local trains,
which operate with a d.c. voltage under 1000 V, is about 25 000 km. These figures show the extent of the d.c.
voltage systems by railways and also the importance of an optimal overvoltage protection which is adjusted to the
specific demands of the d.c. voltage railways.
Increasing use of electronic equipment in and close to the rails and overhead lines (safety and signaling
equipment) need protection against overvoltages. Further on each breakdown of the power supply leads to an
interruption of the train service.
Lightning strokes are the most dangerous threats for railway networks. Overhead lines and trains can be hit by
direct or nearby lightning. For this reason very high charges can be transferred into the overhead lines and the
installed surge protective devices have to withstand high energy stresses. On the other hand the continuous
voltage in d.c. traction systems is naturally a d.c. voltage, which means that all surge protective devices have to be
designed for d.c. application. For this reason MO surge arresters without gaps are used in the d.c. power supply of
the traction systems. It stands for itself that the used MO surge arresters have to be long term stable under d.c.
voltage stress.
The application and dimensioning of metal oxide surge arresters (MO surge arresters) without spark-gaps in
alternating current networks with 50/60 Hz and 16,7 Hz of the railway supply is not very different from the one of
the general energy supply. Requirements and tests for MO surge arresters for application in a.c. traction systems
are similar to the ones for MO surge arresters without gaps for three phase power systems and [IEC 2009] applies.
A separate international standard is not available.
As can be seen in Table 1.2 the voltages in traction systems have a strong fluctuation depending on the load in the
system, which is given by the number of accelerating and breaking trains in a power supply section. Due to this fact
the voltage Umax2 should be considered the relevant precondition when choosing the continuous voltage Uc. This
applies for the a.c. as well as for the d.c. system.
In general the electrical and mechanical requirements for MO surge arresters for application in traction systems are
very high. Arresters installed on traction vehicles have to withstand high mechanical stresses, especially vibrations,
mechanical shocks and high wind loads in case of high speed trains have to be considered.
Because rails, trains and train stations are public places the safety of the surge arresters is an important point. The
arresters should have fail-safe performance in case of an overload.
The electrical requirements and tests for MO surge arresters for application in d.c. traction systems are given in the
new European Standard EN 50526-1 [EN 2012].
Page 24
Nominal voltage
Un
V
Highest permanent
voltage
Highest non-permanent
voltage
Umax1
Umax2
Umax3
V
720
770
1015
750
900
950
1269
1500
1800
1950
2538
3000
3600
3900
5075
17 250
18 000
24 311
27 500
29 000
38 746
Figure 1.4: Highest values of voltage occurring in the system depending on time duration
Page 25
Un :
designated voltage for a system.
Umax1: maximum value of the voltage likely to be present indefinitely.
Umax2: maximum value of the voltage likely to be present for maximum 5 min.
Figure 1.5 shows as an example the possible voltage fluctuations on the current collector of a metropolitan d.c.
traction system. For the purpose of overvoltage protection only the maximum values are of importance, because
they designate the values for the design of the MO surge arresters used for the protection in traction systems.
Figure 1.5: Voltage at the current collector of a d.c. traction vehicle for a period of 15 min, urban
transportation system
1.4.3 MO SURGE ARRESTERS FOR D.C. TRACTION SYSTEMS
The surge arresters are classified by their charge transfer capability Qt and their nominal discharge current In.
The classes DC-A, DC-B and DC-C correspond to increasing discharge requirements. The selection of the
appropriate class shall be based on system requirements.
Class DC-A has a nominal current of In = 10 kA and a charge transfer capability of Qt = 1.0 As
Class DC-B has a nominal current of In = 10 kA and a charge transfer capability of Qt = 2.5 As
Class DC-C has a nominal current of In = 20 kA and a charge transfer capability of Qt = 7.5 As
An optional test is intended to prove the ability of the arrester to withstand direct lightning currents. The
requirements for the direct lightning current Iimp are 2 kA peak value for class DC-A, 5 kA peak value for class DC-B
and 15 kA peak value for the class DC-C.
Page 26
1.5.1 INTRODUCTION
Lightning and thunderstorms occur all over the world, from far north in Norway to far south in South-Africa. The
heaviest thunderstorms with the most intense lightning will normally be experienced in an area of about 2000 km
along the equator.
Lightning has always been a problem to telecommunication and electricity systems and surge arresters have
become an important asset to protect people and equipment against dangerous over-voltages caused by lightning.
As a general rule surge arresters are installed close to the equipment it shall protect.
During a thunder storm a power line could probably be hit by several lightning strokes, either directly to the phase
conductors or most likely to the shielding wires, if any present.
The strokes will cause an earth fault on the line, and initiate a switching sequence of the circuit breakers in the
substations at both line ends.
The following lightning surges and switching surges will eventually stress the surge arresters at the line ends or
elsewhere in the substations.
The degree of stress to the arrester depends on several factors like: distance to the place of lightning stroke,
striking point on the voltage curve, lightning current amplitude, earth resistance, the total flash charge, tower
earthing impedance and if the power line has shielding wires or not.
Different electricity systems with different voltage levels will see different levels of lightning surges and cause
different level of stress to the arresters.
52kV)
Power distribution lines are generally of lower height and less exposed to direct flashes than transmission lines with
higher voltage. The substations are often in-house and well protected against direct strokes to the bus bars. The
number of overvoltages exceeding the basic insulation level in these systems is dominated by induced
overvoltages caused by lightning strokes to or in the surroundings. Due to a large number of surge arresters used
in these systems the energy is split on several units and failures due to stress caused by lightning are rare. Surge
arresters for these systems are rather small and cheap and are easy to keep as spares and to replace when
necessary.
245kV)
Systems within these voltage levels consists of both distribution and transmission lines but often still in a rural area
where the lines and substations to a certain degree are protected against lightning by surrounded houses, towers
and trees. Combinations of surge arrester stress due to direct strokes, back-flashovers and induced voltages will
statistically result in a higher failure rate caused by lightning than in any other systems.
Transmission lines with steel towers and shield wires are in spite of the height above ground well protected against
direct lightning strokes to the phase lines. Most of the lightning will hit the towers or the shield wires, and only a
back-flashover will cause a critical surge in the phase-line.
Page 27
Operation experience from different utilities in different places on earth implies surge arrester failure due to
lightning only in cases with very close and nearby strokes to the substation or in an incoming transmission line
without shielding wires. The surge arrester energy capacity is dimensioned according to the much higher energy in
switching surges and normally the much lower stress from lightning is not a problem.
System voltage Us
kV
0.23 - 52
52 - 245
Insulation characteristic
BIL < induced lightning
surges.
Determined by lightning
overvoltages
BIL determined by
lightning overvoltages.
Lightning overvoltage
Induced surges can cause
flashover.
Direct stroke to a line or
substation is very rare.
Induced surges can cause
flashover.
Direct stroke to shield
wires with back flashover,
and direct stroke to phase
conductors.
Direct stroke to the power
line, to a phase conductor
or shield wire.
2,20 kJ / kV
0,02 kJ / kV
0,03 kJ / kV
According to [Fuk 1997] only lightning strokes hitting the line without shielding wires within 2 km to the substation
will cause critical level of stress to the arresters, about 60 kJ/kV.
In power systems with voltage > 100kV the modern metal oxide surge arrester will be able to withstand the stress
caused by lightning as long as it is not a lightning stroke directly or close to the arrester in a power line without
shielding wires.
0.88kJ/kV
0.04kJ/kV
15.8kJ/kV
2.95 kJ/kV
Page 28
0.10 kJ/kV
Area/Country
Maximum flash
density
50% value
positive
50% value
negative
50% value
total
23
17
18
0.6
37
22
24
24
(51kA 16% value)
0.1 - 200
35
Norway
Norway
west coast
Japan (winter)
1.5.5.1 General
Norway as well as Japan experience rather often thunder storms during winter.
In Norway the winter lightning occurs most frequent along the west coast and not very far inland from the sea.
Typical weather conditions to create the thunder storms are strong winds from the west which bring rather warm air
from the ocean in to the Norwegian mainland. The warm air is pressed upwards and collides with the much colder
air coming from the mountains and the Norwegian inland. Normally heavy clouds are building up and eventually
initiate thunderstorms. Observations from the Norwegian lightning location and registration system shows a
proportional higher number of positive lightning strokes during the winter storms than during the summer storms.
The positive strokes have in general higher energy and higher lightning current than the negative strokes.
Along the northern part of the west coast the winter lightning occurs more frequent than summer lightning, but in
the southern part the summer lightning is much more common and more frequent. In Norway there are no
indications of more failures in the electricity systems caused by lightning in the winter than in the summer.
Winter lightning studies from the engineering side started in 1978 in Japan, because the winter lightning occurred
primarily along the coast of the Sea of Japan which brought about peculiar lightning faults on EHV transmission
lines. The characteristics of winter lightning are described below, based on measurement results so far and are
compared with summer lightning.
Page 29
The duration of the lightning activity of an individual storm is short (usually less than 30 min) and the
frequency of lightning discharge is very low.
About 30 40% of all ground flashes lower from cloud positive charge. This percentage is remarkably high
compared with that of a few percent for summer thunderclouds.
Ground flashes of more than 300 C are occasionally observed [Hac 2008].
Winter thunderclouds in this area are formed by the advection of Siberian air masses, which are dry polar air
masses, over the relatively warm Sea of Japan.
Figure 1.6: Schematic chart illustrating radar echoes associated with the cycle of a thundercloud
(Chisholm and Renick, 1972); in the lower part of this figure the temporal variation of each radar echo
reflectivity in a convective cloud is illustrated in correspondence with the echo life stages.
1.5.5.3 Lightning current parameters
Table 1.5 summarizes typical lightning current parameters in Japan. As is widely known around the world, lightning
in Japan can be broadly classified into summer lightning, which occurs frequently in the summer months, both in
the mountains and flatlands, and winter lightning, which occurs frequently in the winter months, primarily along the
coast of the Sea of Japan. Consequently, the Lightning Protection Design Guide for Transmission Line must
propose designs that take into account the characteristics of both types of lightning.
Page 30
Table 1.5: Characteristic comparison between summer and winter lightning [AIE 1950, Uma,
Uma 1987]
1.5.5.4 Probability distribution of lightning stroke peak current
For the cumulative probability distribution of lightning stroke peak current, a variety of them have been proposed
mainly targeting at summer lightning, both in Japan and abroad CRI 1976 . In Figure 1.7, curve 1 is the one
recommended in the old guide [AIE 1950] (logarithmic normal type: Average value=26kA and logI=0.325), while
curve 5 is an AIEE distribution curve [Uma ] (exponential type). In addition, curve 2 is a distribution created based
on both positive and negative polarity data from 103 winter lightning strokes recorded between January 1979 and
July 1986 in Kashiwazaki and Fukui on the coast of the Sea of Japan. Note that since the cumulative frequency
distribution of lightning peak current differs by the geographical region, we must consider distributions that are
appropriate for each one.
Based on a comparison of probability distributions, this guide recommends curve 1 given by equation 1.2, which
agrees with the observational results for both summer and winter lightning, as the probability distribution of
lightning stroke peak current.
(equation 1.2)
Note that it is possible to calculate the probability for any current value with the above equation, but we do not
recommend it for portions exceeding the range in Figure 1.7.
Page 31
probability (%)
Figure 1.7: Various data of cumulative probability distribution of lightning stroke peak current
[CRI1976]
The winter lightning phenomenon along the coast of the Sea of Japan occurs, is specific in this region on the globe.
The increase of positive ground flashes and frequent occurrence of upward lightning are considered as main
features of the winter lightning in this region.
Page 32
Summer lightning
probability
probability (%)
Winter lightning
Figure 1.8: Probability of the wave front duration of lightning current [Ike 1981][Asa 1994]
As for the range of wave tail duration of lightning current, it is approximately 10~100 s with the 50% value of
30~50 s [Ike 1981].
The observation result for the winter lightning current in Japan is shown in Figure 1.9 [CRI 1989]. The figure
suggests the following:
-
Page 33
probability (%)
Figure 1.9: Probability of lightning current wave tail duration of winter lightning [CRI 1989]
As understood from the relationship between wave front duration and wave tail duration in Figure 1.10, the longer
the wave front duration, the wave front duration also tends to be longer. According to data of Figure 1.11, which
were measured by Berger [Ber 1975], the 10% value of the wave tail duration of negative first stroke currents
exceeds 150 s. When focused on the positive polarity lightning, the 10% value exceeds 1 ms.
Figure 1.10: Relationship between the wave front and wave tail duration of the lightning current of
winter lightning [CRI 1989]
Page 34
probability (%)
Figure 1.11: Probability of wave tail duration of lightning current [Ber 1975]
: Total
probability
30% : 2.9C
Page 35
Time
Figure 1.13: Example of positive polarity current shape of winter lightning measured at Fukui in Japan in
February 1983 [CRI 1995]
1.5.6.2 Electric charge of lightning currents
The electric charge quantity of lightning is calculated by time integration of the lightning current. An example of the
cumulative probability distribution of the electric charge quantity of the winter lightning current is shown in Figure
1.12 [Miy 1992]. The figure suggests the following:
Approximately 10% of the winter lightning currents have the electric charge quantity exceeding 100 C.
The 50% value of the electric charge quantity of the positive polarity lightning is 20 C, which is 20 times of
the 50% value of negative polarity lightning.
Lightning currents with large electric charge quantity have very long wave tail duration rather than high
crest values. Figure 1.13 shows an example of the current wave shape of positive polarity with duration
exceeding 4 ms observed in winter.
According to electrical charge data of summer lightning shown in Figure 1.14 measured by Berger [Ber
1975], the 50% value of the electrical charge of positive polarity lightning is 80 C, which is 10 times or more
of the 50% value of 7.5 C for the negative polarity lightning.
probability
Electrical charge C
Page 36
Page 37
a) Transportation vibration
Arresters are transported by trucks, trailers, freighters or airplanes and subjected to their inherent frequencies and
shocks due to bad roads, blocks, sudden braking or landing. These loads are usually taken into consideration in
the construction of arresters: e.g. insulators for supporting ZnO elements or bumpers between ZnO elements and
the porcelain wall. In case of polymer arresters where ZnO elements are directly molded, these loads may not be
large.
Some manufacturers evaluate the capability against these loads by the actual transportation test, where a truck or
trailer runs through normal roads, highways and bad roads. The vibration test with simulated waveforms is also
carried out instead of the actual transportation test. Shocks watching labels or meters are sometimes used for
monitoring loads when a high shock is expected during transportation.
b) Suspension vibration
A suspension type arrester may be subjected to a continuous vibration of its natural frequency due to the wind or
the movement of the line. This load may be small for the arrester body because acceleration is not so high and the
natural frequency of the suspension system is lower than that of the arrester, but the long-term reliability of the lead
connecting and arrester suspending parts should be confirmed.
(3) Seismic load
This load should be considered when the arrester is installed in the area where a large earthquake is expected. As
a seismic event is rare to occur, it is unreasonable to expect all of the above static loads to occur simultaneously.
There are three typical guides, IEC, IEEE and JEAG (Japan), which have been published and revised through the
experience of the large earthquakes as shown in Table 1.6.
Page 38
Year
1965-
1970-
1975-
1980-
1985-
1990-
st
JEAG
Static 0.5 G
-5003
1995-
2000-
2005-
Revision
Dynamic 0.3 G
(1980)
st
(1998)
IEEE
Revision
Revision
-693
(1984)
(1997)
(2005)
68-3-3
[Note]
68-2-57
(1991)
1463
62271-2
(1989)
1166
(1996)
(2002)
IEC
(1993)
Tottoriken
Niigata
-Seibu
(1964)
Off-
Nihonkai
Hokaido
Hyogoken
Off-
Miyagi
-Chuubu
-Nanseioki
-Nanbu
Tokachi
(1978)
(1983)
(1993)
(1995)
Earthquake
(2000)
in Japan
Niigata
-Chuuetu
(1968)
(2004)
Earthquake
Sanfernand
Northridge
in US
(1971)
(1994)
Indonesia
Kocaeli,
Other earthquake
Off-
Pakistan
Sumatra
(2005)
Shu-shu
(1999)
(2004)
Table 1.6: Publication and revision history of the guide for seismic test [JEA 1998]
Table 1.7 shows the seismic test conditions of IEC, IEEE and JEAG for arresters, where the acceleration response
of JEAG is the most severe. The concept of the JEAG with resonant 3 cycles sine wave is different from that of IEC
and ANSI with artificial earthquake waves which comply with the required response spectrum (RRS) as explained
below.
Seismic loads of arresters are also strongly dependent on the connecting leads [Oka 1986]. The guides of IEEE
and JEAG require sufficient flexible lead slack that allows for any relative deflection of the equipment that will occur
during an earthquake.
Page 39
Arrester standard
IEC60099-4
IEEE C62.11
JEC-2371
(Circuit breaker)
IEEE-693
JEAG-5003
IEC 60068-3-3
(General)
(General)
IEC 62271-300
Referred guide for seismic test
(General)
90kV Test
Voltage rating
170 kV
Not detailed
( 90kV Analysis)
0.5 G High
Input
acceleration
0.5 G High
0.3 G
0.3 G Moderate
0.25 G Moderate
0.2 G Low
Frequency
range
Seismic test
method
0.5 35 Hz
0.3 33 Hz
0.5 10 Hz
Artificial earthquake
which complies with
RRS
Artificial earthquake
which complies with
RRS
Resonant 3 cycles
(Note 2)
sine wave
Waveform
RRS
Max. acceleration response
at 2% damping
RRS
1.4G (High)
1.62G (High)
2.35G
0.85G (Moderate)
(Single-degree-of-freedomsystem)
Acceptance
criteria
For structural
parts
For arrester
performance
0.81G (Moderate)
0.56G (Low)
Specified
Specified
Specified
in the guide
in the guide
in the guide
Not detailed
Not detailed
Not detailed
(Note 1)
(Note 1) Only check items (reference voltage, partial discharge, leakage check) are listed in Annex M of IEC
60099-4.
(Note 2) When the inherent natural frequency of the equipment is higher (lower) than 10 Hz (0.5 Hz), the frequency
of 10 Hz (0.5 Hz) shall be used.
Page 40
Figure 1.15: Distribution of horizontal acceleration on the ground surface expected once at the interval
of 75 years [JEA 1998]
Table 1.7 shows the seismic intensity scale used by the Japanese Meteorological Agency. The acceleration of 0.3
G (294 Gal) corresponds to the level of the seismic intensity .
Figure 1.16 shows the seismic records in the past 75 years from 1921 to 1995 in Japan. This figure shows that the
past records of the acceleration are below 0.3 G (294 Gal) in almost all regions in Japan. Please note that the
mentioned scale was revised in 1997.
Page 41
Intensity scale
Designation
Acceleration (Gal)
No feeling
< 0.8
A slight earthquake
0.8
2.5
A weak earthquake
2.5
A strong earthquake
25
80
80
250
A disastrous earthquake
250
> 400
25
400
Table 1.8: Seismic intensity scale used in 1949 1997 by the Japanese Meteorological Agency
[JEA 1998]
Hokkaido
107
12
Tohoku
116
25
Hokuriku
17
Kanto
13
Chubu
93
Chugoku
205
3
32
Kansai
95
17
Chugoku
58
13
Kyushu
4
Figure 1.16: Seismic records in the past 75 years from 1921 to 1995 in Japan [JEA 1998]
Page 42
(2) Waveform
As the waveform of an earthquake is dependent on the ground condition between the epicenter and the equipment,
it is not practical to specify an earthquake waveform for the test. Porcelain type apparatuses are destroyed at the
peak acceleration of the response and the destruction is not affected by the duration and the waveform of the
vibration, therefore the quasi-resonant method by resonant N cycle sine wave is adopted as the Japanese seismic
test.
(2.1) Frequency
The predominant frequencies of earthquakes in Japan are 0.510 Hz, which overlaps the natural frequencies of
porcelain type apparatuses, bushings, arresters, etc. So the severest test condition is realized when the test wave
is specified as a sine wave with the natural frequency of the apparatus.
(2.2) Cycle
The acceleration response factors of a simplified single-degree-of-freedom model to resonant 13 cycles sine
waves are compared with the actual 615 earthquake records on the ground surface in Figure 1.17. The response
factor in the resonant 2 cycles sine wave covers the most responses of the actual earthquake records.
In addition, the amplification effects due to the existence of foundations (1.2 times) and other unknown factors (1.1
times) are considered. As the correction factor of 1.3 (1.2x1.1) corresponds to the ratio of the response factor to
resonant 3 cycles sine wave to that to resonant 2 cycles sine wave (6.1/4.7 as shown in Figure 1.16), resonant 3
cycles sine wave is adopted as the seismic test wave.
Figure 1.17: Comparison of the response factors between a simplified single-degree model to resonant
1-3 cycles sine waves and the actual 615 earthquake records in Japan [JEA 1998]
Page 43
Direction
Item
Natural frequency [Hz]
Seismic test
JEAG-5003
of
24
17
> 30
Wave shape
0.3
0.31
0.31
Tank (A4)
1.4
1.4
1.0
2.3
2.9
1.0
2.4
3.0
1.1
Response
factor
Wave shape
0.35
0.35
---
Tank (A4)
1.1
1.0
---
1.3
1.0
---
1.3
1.1
---
the EL Centro
Response
earthquake wave
factor
Table 1.9: Seismic test results of 500 kV GIS arrester in accordance with JEAG-5003 [Shi 1997]
Page 44
Tank (A4)
Middle of internal parts (A3)
Top of internal parts (A2)
Input acceleration (A1)
Time
a: Example of wave shapes
Tank (A4)
X
Z
Y
1.5 m
Top of
internal parts (A2)
Middle of
internal parts (A3)
Shield
ZnO element
Shaking table
b: 500 kV GIS shaking table,
seismic test for horizontal
installation
X
Z
Figure 1.18: Seismic test of 500 kV GIS arrester in accordance with JEAG-5003 [Shi1997]. The axis X,Y
and Z give the directions of acceleration, see Table 1.9.
Page 45
1.6.2 POLLUTION
Authors in charge: Bernhard Richter and Yoshihiro Ishizaki
One of the unsolved problems, and therefore discussed in Cigr working groups again, is the pollution performance
of polymer housed surge arresters for high voltage applications. Due to the length of the multi-unit designs test
procedures and criteria could not be agreed upon in the past. This was discussed as well in WG A3.21 and WG
A3.22 when dealing with 1000 kV UHV aspects. The principle problems are addressed in the following.
Figure 1.19 illustrates three possible mechanisms that may affect a multi-unit MO HV arrester in a polluted
environment, see also [Ric 2007].
A special problem for HV arresters of Type A (see 2.3 Design of surge arresters) with considerable gas in the
inside of the insulator may be the radial field strength as shown in Figure 1.20. Radial field stress, however, is also
a concern for Type B arresters. Here the radial voltage drops across small distance of only few millimeters between
the outer surface and the MO column, and a weak design may lead to puncture of the insulation material. As the
possible radial field stress increases with the distance between two metal fittings, the maximum unit length is
limited.
3
Risk of "internal"
partial discharges,
degradation of the
MO resistors and
deterioration of the
supporting structure
1
Risk of external
flashover (see
IEC 60507)
2
Risk of partial
heating of the active
parts (see Annex F
of IEC 60099-4)
Page 46
MO column
Gas
Conductive
layer
Uaxial, int
Solid
Uradial
Figure 1.20: Possible voltage distribution of an arrester unit under polluted conditions
Table 1.10 gives an overview about proposed methods for pollution tests of polymer housed MO surge
arresters. The main discussed and open point is the way of treating the polymer surface of the insulating
housing to get realistic results in the test, which can be compared to long term stresses in the system.
Item
Solid layer
Standard to be
referred
IEC 60507
Artificial
pollution
procedure
Pollutant slurry
sprayed on insulator
Soluble
component
NaCl
NaCl
NaCl
NaCl
Kieselguhr
Kaolin
Tonoko
SiO2
or Tonoko
or Bentonite
Non-solible
component
Note
Sprayed pollutant
Salt fog
JEC 0201
IEC 60507
IEEE C62.11
--Recovery of hydrophobicity
during testing considered
No special testing
facilities required
Recovery of
hydrophobicity
considered
Page 47
Remark 1: Tonoko
Tonoko is fine inorganic powder from particular kinds of natural rock, which has been originally used as fine
abrasive or filler for some traditional craft works in Japan. Tonoko is also introduced in IEC 60507 as inert material,
of which characteristics are given in Table 1.11.
Granulometry
(cumulative distribution) mm
SiO2
Al2O3
Fe2O3
H2O
16%
50%
84%
Volume
conductivity
s20 (S/m)
60 - 70
10 20
4-8
0.8 - 1.5
35
8 - 15
0.002-0.01
Figure 1.21: Procedure to remove hydrophobicity and to apply contaminant for solid layer method
[Nai1996][Ish2008]
Remark 2: representative methods to remove hydrophobicity
The following procedure, as shown in Figure 1.21, is suggested to remove hydrophobicity on silicone surface of
MOR temporarily for the testing, without any damage on the surface or any additional chemical agent in the
pollutant.
a) Prepare Tonoko slurry, which contains approx. 1 kg of Tonoko in 1 liter of water.
b) Spray the Tonoko slurry as uniformly as possible on the hydrophobic MOR surface.
c) Dry the polluted surface under natural ambient condition.
Page 48
d) Wash off the deposited Tonoko roughly, by running tap water, for example. After this process some amount of
Tonoko will remain on the surface, which suppresses recovery of the hydrophobicity temporarily. It is important
to conduct testing during hydrophobicity is completely lost.
Voltage application method
Figure 1.22 shows the test with only the MCOV to apply to surge arrester under the polluted condition according
with IEC 60099-4, Annex F.
Figure 1.23 shows the voltage application method to superimpose a temporary overvoltage (the sound phase
voltage at the time of single phase ground fault) on the continuous operating voltage (U c) according to JEC
standard. This test procedure is considered as one of solutions for the risk concerned the external flashover at TOV
of surge arresters.
10 min.
contaminant
application
UC
3 min.
Figure 1.22: Continuous voltage U c applying method according to IEC standard [IEC 2006]
4 sequences
E2
E1
1 min
contaminant
application
3 min
1 min
1 min
1 min
1 min
E1
E2
1 min
1 min
1 min
first sequence
Figure 1.23: Temporary overvoltage (TOV) applying method according to JEC standard
[JEC 2003]
Page 49
1.6.3 HUMIDITY
Author in charge: Kari Lahti
Figure 1.24: The worlds climatic zones. The most humid climates in equatorial and tropical zones are
indicated by colors [Mar92]
Page 50
Ng
0.04TD1.25
-2
(equation 1.3)
-1
Page 51
Group
Arrester types
Description
A,B,C,H
II
D,E
III
F,G
Leakage Current ( A)
Table 1.12: Tested medium voltage arrester types according to internal structure
100
90
80
70
60
50
40
30
20
Group I
Group II
Group III
10
0
0
100
200
300
400
500
600
700
800
Figure 1.26: internal leakage currents of MO surge arresters according to their internal structure during
the humidity chamber test
Page 52
Based on the results, at least slight moisture penetration into arresters interior is possible in most of the arrester
types studied. However, remarkable levels of internal leakage current was reached by all the arresters of type III,
most of the arresters of type II but none of the arresters of type I. In this context one has to keep in mind that these
tests are performed for certain MV arrester types and, for example, in HV arresters also clearly different structures
are utilized.
In general, same kinds of results/behaviors have been achieved also in hot water immersion tests [Lah 1999] [Lah
2003]. Based on this result hot water immersion tests are applicable in sealing testing of polymer housed arresters.
In some cases some problems may be found in hot water immersion testing of arresters with housings with good
surface properties and high diffusion coefficient (e.g. some silicones). Quite high internal power losses may be
measured immediately after immersion test, which result does not necessarily reflect real service situation where
also good surface properties (e.g. hydrophobicity) affects the total behavior.
Page 53
10
9
D1
D2
D3
D4
D5
D6
8
Power losses (W)
7
6
3
2
1
0
0
10
20
30
40
50
60
Test duration (weeks)
70
80
90
100
Figure 1.27: Internal power losses of type D (internal structure type II) at 12 kV during the test
E1
E2
E3
E4
E5
E6
E7
F1
10
9
8
Power losses (W)
7
6
5
4
3
2
1
0
0
10
20
30
40
50
60
Test duration (weeks)
70
80
90
100
Figure 1.28: Internal power losses of type E (internal structure type II) and F (internal structure type III)
at 12 kV during the test
5
H1
H2
H3
H4
A1
A2
4
3
2
1
0
0
10
20
30
40
50
60
Test duration (weeks)
70
80
90
100
Figure 1.29: Internal power losses of type A and H (both of internal structure type I) at 12 kV during
test
Page 54
Page 55
Page 56
Page 57
Page 58
Figure 1.32: Short circuit behavior of a cage design polymer housed arrester unit (courtesy ABB).
1: the arrester has failed and gas begins to be expelled through the housing
2: the gas streams trigger an external flashover and the internal arc is commutated to the outside
The maximum short circuit current, defined for a flowing time of 200 ms, is the rated short circuit current Is, given as
symmetric current in r.m.s. value with power frequency. The manufacturer has to state the rated short circuit
current the arrester is intended to withstand. Table 1.13 gives the rated short circuit currents for short circuit tests
as they are defined in IEC 60099-4.
600 200
Table 1.13: Standard short circuit currents for tests purposes of MO arresters
Once the rated short circuit current is claimed for an MO arrester the type tests have to be performed with all the
short circuit currents given in Table 1.13. The required tests at reduced short circuit currents shall help avoiding
that the short circuit performance is optimized for high short circuit currents only while at lower short circuit currents
the housing might violently shatter before the pressure relief devices can open.
Page 59
To avoid misunderstandings, it has to be pointed out that the short circuit current with power frequency, driven by
the system voltage, is flowing through the surge arrester, or more likely through the arc, only after the arrester was
overloaded. Therefore, the MO arrester was destroyed and failed in a controlled way.
Figure 1.33 shows as example porcelain housed MO arrester units after successful short circuit tests. Both the
results are considered to be positive according the pass criteria as defined in the test standard IEC 60099-4.
Thermal cracking, as to be seen in Figure 1.33 (right) is acceptable, as long as all parts heavier than 60 g remain in
a well-defined enclosure. The MO arrester is of course destroyed in both cases, but it failed in a well-defined
manner.
Figure 1.33: Examples for successful short circuit tests of porcelain housed arrester units.
left: intact porcelain body, right: secondary break of porcelain body(courtesy Siemens).
Rated short circuit current tested I s = 63 kA for 200 ms.
Figure 1.34 shows examples of short circuit tests performed on polymer housed MO arrester units. Both results are
to be considered positive. In both cases the arrester units are destroyed, but without cracking or ejecting hard parts
like MO resistors or parts of them. In case of the cage design, Figure 1.34 right, only soft polymeric material is
ripped off.
Page 60
Figure 1.34: Example of successful short circuit tests on polymer housed MO arrester units.
left: FRP hollow insulator (courtesy Siemens),
right: cage design (courtesy ABB).
In both cases the short circuit current was I s = 63 kA for 200 ms.
It has to be understood clearly that an overload is a normal operation for a surge arrester. Therefore, surge
arresters have to be designed for an overload. Important is that the arrester fails in a controlled and safe way. This
is an important safety issue when arresters are installed close to public places, sensitive infrastructure or for
instance on the roof of a traction vehicle in railway applications.
Page 61
2.
2.1.1 INTRODUCTION
A metal-oxide surge arrester without gaps (MO arrester) is an arrester having nonlinear metal-oxide resistors
connected in series and/or in parallel without any integrated series or parallel spark gaps. The wording surge
arrester is used in the HV and MV community and describes different designs of MO arresters. In the LV field it is
common to speak in general about Surge Protective Devices (SPDs), which covers different technologies and
design types, e.g. spark gaps, metal oxide varistors and combinations of them including disconnecting devices etc.
The function of a surge arrester with an active part consisting of a series connection of MO resistors is very simple.
In the event of a voltage increase at the arresters terminals, the current rises according to the characteristic curve,
see Figure 2.1, continually and without delay, which means that there is no actual spark over, but that the arrester
skips over to the conducting condition. After the overvoltage subsides the current becomes smaller according to the
characteristic curve. A subsequent current, such as those that arise with spark-gaps and spark-gapped arresters,
does not exist; it flows only the so-called almost pure capacitive leakage current ic of about 1mA.
Page 62
In Figure 2.2 a more technical diagram is given, indicating the standardised definitions.
Figure 2.2: Voltage-current characteristic of a MO arrester with I n = 10 kA, line discharge class 2. The
voltage is normalized to the residual voltage of the arrester at I n . The values are given as peak values
for the voltage (linear scale) and the current (logarithmic scale). Shown are typical values.
In Figure 2.3 typical voltage and current wave forms are given for MO surge arrester as defined in IEC 60099-4 and
ANSI/IEEE C62.11. In the lower part at Uc the arrester acts as a capacitor, the current is in the range of 1 mA and
below. At the knee point b (Figure 2.1) at Ur and Uref the arrester starts to conduct, the ohmic content of the current
is increasing rapidly with a slight voltage increase. At U ref the current has a dominantly ohmic component.
Temporary overvoltages have to be considered in the voltage region between points b and c.
In the low current region up to point b power frequency currents and voltages have to be considered (range of
continuous operation). In the region above b the protective characteristic of the MO arrester is of importance.
Thats why in this range the voltage-current-characteristic is defined by impulse currents of different wave shapes
and current magnitudes (Figure 2.2).
For simulations of the performance of MO surge arresters normally a voltage-current-characteristic is used that
starts at some 10 A in the region c and goes up to maximal 40 kA in the region d. Normally an impulse current
wave shape of 8/20 s is considered.
Page 63
Uc
Isw
Ur
In
Uref
Ist
Ihc
Uc
Ur
Uref
Isw
In
Ist
Ihc
The following paragraph briefly explains typical current and voltage waveforms in different areas of the
characteristic curve.
Page 64
Reference voltage Uref: Peak value of the power-frequency voltage divided by 2, which is applied to the arrester
to obtain the reference current.
Reference current iref: Peak value (the higher peak value of the two polarities if the current is asymmetrical) of the
resistive component of a power-frequency current used to determine the reference voltage of an arrester.
The reference current is chosen by the manufacturer in such a way that it lies above the knee point of the voltagecurrent characteristic and has a dominant ohmic component. Therefore, the influences of the stray capacitance of
the arrester at the measurement of the reference voltage are not to be taken into account. The reference voltages,
which are measured at single MO resistors, can be added to give the reference voltage of the entire arrester.
Reference voltage U1mA and reference current with DC voltage:
A reference current and the reference voltage for DC voltage belonging to it are often also demanded instead of a
given reference current for AC voltage. It is now common practice to specify the DC voltage, which is applied with a
direct current of 1 mA to the terminals, no matter what the diameters of the MO resistors are. Both types of
information, the reference current and the reference voltage for AC voltage and for DC voltage, are in principle
equal. Both of these types information describe a point on the voltage-current characteristic of an arrester, where
the influences of the stray capacitance can be ignored. All the tests performed according to IEC are always based
on the reference current and the reference voltage for AC voltage. Reference current and reference voltage with
DC voltage are additional information, which can be received from the manufacturer.
Residual voltage Ures: Peak value of voltage that appears between the arrester terminals during the passage of a
discharge current.
The residual voltage of a MO resistor or MO arrester is determined with surges having different wave forms and
current heights and it is given in tables or as a voltage-current characteristic on a curve. The measurements are
generally performed on MO resistors. As the measurement is mostly performed in regions of the characteristic
where the ohmic part of the current is dominant, the capacitive stray influences can be ignored. The residual
voltages measured on single MO resistors can be summed up as the residual voltages of the whole arrester.
Lightning impulse protective level Upl: Maximum permissible peak voltage on the terminals of an arrester
subjected to the nominal discharge current. Corresponds to the guaranteed residual voltage Ures at In.
Switching impulse protective level Ups: Maximum permissible peak value on the terminals of an arrester
subjected to switching impulses.
Lightning current impulse: Current impulse with the wave shape 8/20 s. The virtual front time is 8 s and the
time to half-value on the tail is 20 s. The lightning current impulse reproduces approximately the current impulse
produced by a lightning stroke in a conductor after an insulator flashover. This current impulse travels as a
transient wave along the line.
Nominal discharge current of an arrester In: The peak value of the lightning current impulse that is used to
classify an arrester. The nominal discharge current and the line discharge class of an arrester are correlated to the
system voltages and prescribe the test parameters, see Table 2.1. Recommendations for the choice of the nominal
discharge currents and the line discharge classes for different system voltages are to be found in IEC 2000 and
IEC 2009 .
High current impulse Ihc: Peak value of discharge current having a 4/10 s impulse shape. The high current
impulse should reproduce a lightning stroke close to an arrester and it is used with medium voltage arresters of the
line discharge class 1 as a proof of thermal stability. It represents not only an energetic stress but also a dielectric
one, taking into consideration the high residual voltage that occurs with a high current impulse with a peak value of
100 kA. It is however, necessary to strongly emphasize that a high current impulse with an amplitude of 100 kA is
not the same as a real lightning current of the same amplitude. The real lightning current of this amplitude
measured during a thunderstorm lasts longer than several 100 s. Though such strong lightning currents and
impulse shapes are very rare and appear only under special conditions, such as during winter lightning in hilly
coastal areas.
Page 65
Switching current impulse Isw: Peak value of discharge current with a virtual front time between 30 s and 100
s, and a virtual time to half-value on the tail of roughly twice the virtual front time. The switching current impulses
are used to determine the voltage-current characteristic, and in connection with the line discharge class are also
used to determine the energy which must be absorbed during the operation test. The current amplitudes lie
between 125 A and 2 kA, and roughly reproduce the load of an arrester produced by overvoltages, which were
caused by circuit breaker operation.
Steep current impulse: Current impulse with a virtual front time of 1 s and a virtual time to half-value on the tail
not longer than 20 s. The steep current impulses are used to determine the voltage-current characteristic. They
have amplitudes up to 20 kA and roughly reproduce steep current impulses like those which may appear with
disconnector operation, re-striking, back flashes, and vacuum circuit breakers.
All the current impulses described above (except the high current impulse) are used to determine the voltagecurrent characteristic of a MO arrester. It must be considered that only the virtual front time and the amplitude of
the current impulses are decisive for the residual voltage and not the virtual time to half-value on the tail. That is the
reason why the tolerance for the virtual front times is very tight, and contrastingly, the tolerances for the virtual
times to half-value on the tail are very broad.
Long-duration current impulse Ild: Also called rectangular wave (Irw) or square wave, a long-duration current
impulse is a rectangular impulse that rises rapidly to its peak value and remains constant for a specified period of
time before it falls rapidly to zero. The length of the current pulse duration is correlated to the line discharge class
of an arrester. Rectangular impulses are used in laboratories during the type tests with long-duration current
impulses, and during the operating duty test of MO arresters having line discharge classes 2 to 5, in order to inject
the energy in the arrester. The current amplitudes are up to 2 kA and reproduce the load of an arrester when a
charged transmission line discharges into the arrester in case of an overvoltage occurrence.
It is now regarded as a matter of course to use a rectangular wave of 2 ms duration to compare different MO
arresters, although there is no norm established for doing so. Specified is either the amplitude of the rectangular
wave for a specific MO arrester or the energy transferred into the arrester during the flow of the rectangular current.
Line discharge class: The line discharge class is the only possible way to specify the energy absorption capability
of an arrester provided in IEC 60099-4. The line discharge classes 1 to 5 are defined with growing demands. They
differ from one another due to the test parameters of the line discharge tests. The energy W is calculated from the
line discharge class in connection with the residual voltage of the switching current impulse. This calculated energy
must be injected with each discharge in a MO resistor during the test with a long-duration current impulse Ild (line
discharge test). Two corresponding line discharges are loaded in the arrester during the operating duty test as a
proof of thermal stability.
W = Ures
(UL Ures)
1/Z
(equation 2.1)
Ures = Residual voltage of the switching current impulse. Here, Ures is the lowest value of the residual voltage
measured at the test sample with the lower value of the switching current impulse.
UL = Charging voltage of the current impulse generator used in test labs for producing the long-duration current
impulse Ild.
Z = Surge impedance current impulse generator.
T = Duration of the long-duration current impulse
The parameter of the line discharge classes are derived from the stored energy of long transmission lines, see
Table 2.1.
That is the reason why the line discharge classes have no direct importance in medium voltage systems. They
serve here only to distinguish the energy handling capability of different arresters.
Page 66
In
LD
kA
Us
kV
km
ZL
T
ms
10
245
300
450
2.0
10
300
300
400
2.0
10
420
360
350
2.4
20
525
420
325
2.8
20
765
480
300
3.2
Table 2.1: Correlation between line discharge classes and parameters of transmission lines. The
duration T of the long-duration current impulse I l d is also given.
Rated short circuit current Is: The r.m.s. value of the highest symmetrical short circuit current, which can flow
after an overload of the arrester through the arc short circuiting the MO resistors without violent shattering of the
housing. The proof of the specified value specified by the manufacturer is conducted in the short circuit test.
It has to limit the occurring overvoltage to a value that is not critical for the electrical equipment and
It has to guarantee a safe and reliable service in the system.
The continuous operating voltage Uc is to be chosen in such a way that the arrester can withstand all power
frequency voltages and also temporary overvoltages without being overloaded in any possible situation. This
means that T Uc must be always higher than the maximum possible temporary overvoltages UTOV in the system.
Comment: Ferromagnetic resonances are the exception. The ferromagnetic resonances can become so high and
exist so long that they may not be taken into consideration by the dimensioning of the continuous voltage if the
arrester should still be able to fulfill its protection function in a meaningful way. If ferromagnetic resonances appear,
then this generally means that the arrester is overloaded. The system user should take the necessary measures to
avoid ferromagnetic resonances.
The MO arrester can fulfill its function of protection properly if the lightning impulse protection level Upl lies clearly
below the lightning impulse withstanding voltage (LIWV) of the electrical equipment to be protected, the safety
factor Ks is also to be taken into consideration. The point is to set the voltage-current characteristic of the arrester in
a way that both requirements are met.
Page 67
It makes sense to choose the continuous operating voltage Uc a little bit higher than was calculated (for instance
10%). As a rule, there is enough distance between the maximum admissible voltage at the electrical equipment
and the protection level of the arrester.
Figure 2.4: Comparison of the possible occurring voltages in the system, the withstand voltage of the
electrical equipment and the parameters of the MO arresters.
Page 68
A Pre-breakdown region
B Breakdown region
C Upturn region
1 DC voltage characteristic
2 AC voltage characteristic
3 Residual voltage characteristic
E Field strength
J Current density
UG Continuous operating voltage (DC)
UB Breakdown (or switching) voltage
U v Continuous operating voltage (AC, 50 Hz)
Up Residual voltage, 8/20 s
Resistivity
Non-linearity exponent (U)
Page 69
By careful dimensioning of the geometry and controlled manufacture of the metal-oxide ceramic, the breakdown
4
voltage per element can be set to values within a very wide range (UB 3 V to >10 V). This allows realizing
protection devices for electronic circuits up to ultra-high voltage systems.
The switching mechanism of the material can be traced back to individual grain boundaries in the ceramics, which
each show a typical value for UB of app.3.2-3.4 V. For a general review see [Lev 1989], [Gre 1990], [Cla 1999],
[Bue 2008]. Combining many grain boundaries in series and in parallel within a MO-element then allows scaling the
voltage and current characteristic of a MO-ceramic block. For a sufficient number of grain boundaries the field
strength E and current density J then describe the material characteristic more generally. In the log-log
representation of the characteristic (Figure 2.5b) there are three distinct regions, i.e. the pre-breakdown region A,
the breakdown regime B, and the upturn region C.
During normal system operation, in which no overvoltages occur, the voltage applied to the arrester is the
continuous operating value (UV or COV for AC or UG for DC), which lies in the upper part of the pre-breakdown
region. The breakdown region is characterized by a very high non-linearity in the current-voltage curve. As Figure
2.5b shows, it is described quantitatively by the non-linearity exponent (E), which is a function of the voltage U or
applied field E, respectively. Its maximum values are typically around 20-70, but values above 100 have been
observed. The rated voltage UR, which also dictates the arrester's range of application, lies in the region of UB. In
the extreme case, where current densities are very high or very low, approaches unity (ohmic behavior), the two
regions being as much as 12 current decades apart [Gre 1989] [Per 2003]! In the case of a high, transient
overvoltage, the varistor state lies in the upper part of the breakdown or in the upturn region. This protective range
is characterized by the residual voltage UP, which depends upon the wave shape and amplitude of the impulse
current. For an arrester to effectively suppress voltage transients, the difference between the residual voltage and
the continuous operating or rated voltage must be small. However, it is important to recognize that the highly
temperature and voltage dependent power loss P v(V,T) generated at UV limits the maximum continuous operating
voltage (MCOV) which is possible. The power loss Pv must be low enough to satisfy the conditions for a thermally
stable state under possibly simultaneous conditions of elevated operating voltage, elevated temperature, aging,
pollution and energy absorption. A broad range of thermal stability is ensured by low power losses and the efficient
dissipation of heat in the arrester. This is also dependent on its detailed internal and external design and the
materials used to make them.
The AC-characteristics shown in Figure 2.5b is the result of the superimposed capacitive and resistive currents
flowing through the grain boundaries under a time dependent voltage stress. It is common practice to plot the
current peak vs. the voltage peak under AC. Well below UV the current is predominantly capacitive, while it is
strongly (non-linear) resistive for peak values above U B . To a first approximation, the AC response can be
described by using the DC-curve for the resistive part and the (small signal) dielectric permittivity of the material,
which is rather high (few hundreds). In some cases, for the so called AC resistive currents also the current value
I(t) in phase with the voltage maximum is used, which provides an improved (but not full) approximation to the
power losses. As indicated in Figure 2.5b, in the breakdown region the peak AC voltage is a few percent above the
steady state DC voltage due to the dynamic effects of charge trapping at the grain boundaries [Gre 1990].
It is worth mentioning that the phenomenon of electrical ageing, leading to thermal instability and failure, was a
serious one early in the development of the ZnO varistor. Today it has largely been brought under control by most
qualified manufacturers. However ageing is not only a result of long term exposure to operating voltages, it can
also be generated by current impulses (impulse degradation). Equally, not only the power losses at operating
voltage and/or reference voltage are affected, in some severe cases the discharge voltage can also be in- or
decreased [Lev 1989], [Per 2003].
Page 70
breakdown, breakdown and lower part of the upturn region. By adding a few percent of selected doping elements
such as Bi, Sb, Co, Mn etc. to ZnO and using a suitable sintering process, it is possible to influence both the
conductivity of the ZnO grains and the properties of the high-resistance grain boundaries. The microstructure of the
ceramic is dominated by the densely packed ZnO grains resembling irregular polyhedrons. It are the common
interfaces (or grain boundaries) of these ZnO-polyhedrons, which are the electrical active part of the material.
Photographs of fracture surfaces provide a clear picture of this structure (Figure 2.6).
Figure 2.6: Typical microstructure of a ZnO-varistor after fracturing preferentially along the grain
boundaries (dark: doped ZnO grains, white: Bi2O3-phase at triple points, grey: Spinel secondary phase
Most of the admixed bismuth oxide collects as a separate phase at the triple points at which the adjacent grain
edges make contact. Also found at these points is a spinel phase in the form of fine grains, which are most easily
distinguished from the ZnO grains by their smaller size and more regular, octahedral shape. The grain boundaries
themselves, however, are not quite free of bismuth, although this cannot be seen in Figure 2.6. Using highly
sensitive techniques for analysis, it is however possible to detect at these boundaries the presence of fractions of
an atomic monolayer of both bismuth and oxygen atoms, which are essential for the electrical function of the
varistor [Cla 1999], [Chi 1998], [Kob 1998], [Sat 2007], [Stu 1990], [Elf 2002], [Che 1996]. During the sintering
process the bismuth oxide melts to form a liquid phase, which dissolves, at least in part, the other doping
substances and promotes their uniform distribution. The liquid phase also favors grain growth and dense sintering.
The spinel precipitates, on the other hand, inhibit grain growth and generate a uniform distribution of the ZnO grain
size. More recently the frequently observed inversion (or twin) boundaries were also recognized to be crucial for
the grain growth mechanism [Ber 2008].
Page 71
1
1.
2.
3.
4.
5.
6.
7.
The basic material used to manufacture metal-oxide resistors is very fine grain ZnO with particle sizes of about
1 m, to which as many as ten or more doping elements are added in the form of fine oxide powders. Its actual
composition differs from manufacturer to manufacturer. The proportion by weight of all additives together is typically
10 percent, with the share of the individual components ranging from ppm to percent. The purity and fineness of
the metal-oxide powders and the homogeneity of the mixture are, therefore, of immense importance for the quality
of the end-product. To achieve the required homogeneity the powder is usually treated in several different
processing steps. Sometimes the metal-oxides are wholly or partially heat-treated or calcined with none, part or all
of the ZnO powder to complete some of the solid-state chemical reactions before the sintering process is carried
out. Almost always a grinding operation is necessary to make the overall grain-size distribution smaller and
narrower and thoroughly mix the powders. Additional mixing operations are also used, in particular to mix the
smaller quantity of metal-oxide material with the larger quantity of ZnO. High shear mixing is frequently used to
achieve high homogeneity and various organic processing aids (dispersants, binders etc.) are added.
After these powder processes have been completed, the mixture or slurry has to be spray-dried to remove the
water and obtain a dry granulate that is beneficial for processing. Sometimes the spray-drying operation is also
carried out for the calcining operation mentioned above. The spheroidal granules obtained by spray-drying have
about 100 m in diameter, flow very easily and can be easily compacted under pressure. This takes place in the
next production stage, during which the granulate is compressed into disc-shaped blocks using a dry, uniaxial,
hydraulic press. The green blocks have approximately 50 to 60 percent of the theoretical density at this stage. It is
important here to ensure a uniform high density throughout the block, and that there are no defects present.
The blocks are finally sintered at about 1100-1300C, which has the effect of fully densifying the compacted powder
into a solid ceramic body with virtually no remaining porosity. In a prior step that occurs at lower temperature, the
organic additives mentioned above are pyrolyzed. This is a critical step, which requires very different conditions of
heat treatment and a continuous flow of fresh air. It can easily introduce flaws and voids into the body. During the
Page 72
sintering process the submicron-sized powder particles are united by means of diffusion and grow into large single
crystalline grains, where at the same time the additive dopants are built into the crystal lattices and the grain
boundaries are formed.
Ready-to-use MO resistors are obtained by adding high-conductivity metal contacts to the flat surfaces and
applying a coating to the resistor's peripheral surface to protect it from the environment. Frequently and beneficially
this is a glass coating, but other organic or inorganic materials have also been used. The coating is often referred
to as a passivation, in analogy with those used in solid-state semiconductors. Therefore, whatever material is
used, it is important that the coating not only has high dielectric and thermal withstand capability but also does not
change the properties of the varistor material underneath it.
Page 73
At the grain boundaries thin intergranular films have consistently been observed, which seems to be an equilibrium
feature decorating all ZnO interfaces in the microstructure [Cla 1999], [Chi 1998], [Kob 1998], [Sat 2007], [Stu
1990], [Elf 2002], [Che 1996]. These films have a thickness of 1 nm, consist of an amorphous Bi2O3-ZnO solid
solution of reduced density and their Bi-concentration corresponds to an equivalent of 0.5-1 atomic layer. In
addition an excess of oxygen ( 0.5-1 monolayer) has been detected at these electrically active interfaces and has
been shown to be essential for the electrical characteristics [Stu 1990]. The negative charge trapped in this
intergranular film (or at its interface to the ZnO grains) is compensated by the ionized, positive donors in the
adjacent space charge (or depletion) regions to achieve charge neutrality. It are these tiny electrostatic potential
barriers B , which control the current flow through the material. Note that these barriers are very thin ( 200nm,
compared to the grain size of 10-20 m) and all the voltage drops occur on the positive biased side of the high
resistive grain boundaries, except in the upturn region, where the finite conductivity of the highly doped ZnO-grains
0.1-1 cm) [Lev 1989], [Cab 2004] starts to add to the voltage drop.
For a good varistor characteristics the potential barrier B(U) should stay high with increasing voltage U to prevent
conduction electrons to be thermally activated over the barrier and to generate a leakage current I(U) or power
loss, respectively (note that: I~exp( B(U) + EF)/kT, for eU >>kT, E F : Fermi energy). If the interface states Ni(E)
indicated in Figure 2.9 have a high density and a suitable energy distribution within the band gap of ZnO, then the
barrier height B(U) hardly changes with voltage and the leakage currents stay low. This is referred to as pinning of
the potential barrier B(U) and eventually even -values <1 can thereby result in the pre-breakdown. If all interface
states Ni(E) however would be filled, the total interface charge Qi ( Qi = Ni(E)dE ) would be constant and B(U)
would decay rapidly for U>0, leading to a poor characteristics. If however filled and unfilled interface states are
available, this decay of B(U) is counteracted by extra electrons being trapped in the unfilled states, Qi(U) will
2
increase and stabilize B(U) at a high level (note that: B(U) ~ Qi /N0 , N0: free carrier density in ZnO). Such a
pinning mechanism for interfacial barriers is also well known for other semiconductors [Gre 1990]. From various
studies the picture evolves that these pinning states are characteristic defect levels of the ZnO crystal lattice
(and/or the intergranular film) and hence are always positioned at the same energy within the energy gap of ZnO
[Gre 1990], [Gre 1986], [Gre 1995]. Due to its chemical nature, ZnO never is fully stochiometric and always
contains a small excess (few ppm) of Zn-atoms (e.g. either via Zn-interstitials or O-vacancies in the crystal lattice).
By varying dopants and the sintering and/or heat treatment in the manufacturing process the defect equilibrium of
ZnO is changed and hence the concentration (but not the energy position) of these pinning states is changed. This
affects the voltage response of the barrier B(U) and thereby different J(E) characteristics will result. A high density
of unfilled states (at U=0) is favorable for low leakage currents (and low power losses), whereas a low density
results in a gradual decay of B(U) and in a higher power loss in the pre-breakdown [Gre 1990], Gre 1986].
Quantitatively the defect equilibrium in ZnO-varistors is still badly understood [Mah 1983], [Suk 1988], [Koh 2000],
[Car 2003], but qualitatively this DSB defect model [Gre 1990], [Gre 1986], together with todays knowledge on
the grain boundary films [Chi 1998], explains why so many different material recipes lead to a varistor behavior and
why there are subtle, but technically important differences in the electrical characteristics depending on the recipe
and process used by the manufacturers.
Next we discuss the breakdown region with its high nonlinearities and a well reproducible switching voltage UB of
ca. 3.2-3.4 V per grain boundary, largely independent on the materials formulation. In the past, several models
were able to explain -values up to ca. 20 [Lev 1989], whereas for good varistors -values well above 50 are
realized today. In the literature often the term varistor is used whenever -values >1 are measured for the J(E)
characteristics, although this is misleading for the present topic. There are various mechanisms in solid state
physics leading to deviations from ohmic behavior, but the real challenge is to explain the high nonlinearities >20.
To achieve such high nonlinearities two ingredients are needed:
i) first a strong pinning of
B(U)
stabilizing
at typically
ii) second a mechanism leading to a raid decay of B(U) for U > UB , since for eU >>kT
- (eU/kT)d B/dU, i.e. U and d B/dU both have to be high for a high nonlinearity!
In the case of a strong pinning of the Fermi level (or B), very high electrical fields build-up on the positively biased
17
3
side of the grain boundary. For sufficiently high doping of the bulk of the ZnO ( N0 > ca. 10 /cm ) the fields at the
interface reach values as high as 0.5-1 MV/cm, high enough to create hot electrons similar to what is known for
Page 74
other semiconductors with high carrier mobilitys (e.g. GaAs). These hot electrons can create holes in the valence
band by impact ionization as soon as their energy near the edge of the space charge region is above the band gap
of ZnO (ca. 3.2 eV). The positively charged holes will diffuse back to the grain boundary within less than 1 nsec.
There they compensate part of the negative interface charge Qi . As a consequence this will lower B and increase
the current across the barrier - and hence also the number of hot electrons created will increase further. This hole
production above a threshold level typical for ZnO will trigger a positive feedback mechanism, which leads to a
rapid decay of B(U) with increasing U and hence to high nonlinearities ! For a sufficiently high hole generation
rate, the energy gain ( B+eU) of the electrons must be in the range of 4 eV, which explains the observed switchingor breakdown-voltage of typically 3.2-3.4 V per grain boundary ( B(UB) 0.8 eV, details also depend on N0 and
Ni(E) [Bla 1986]). The stabilizing element in this avalanche-type feedback mechanism is the electron-hole
recombination at the interface.
Figure 2.10: Band diagram scheme of the hole induced breakdown mechanism
[Gre 1989]
Page 75
With this hole induced breakdown model, developed by Pike, Greuter and Blatter [Gre 1990], [Bla 1986], [Pik
1984], most of the unusual breakdown phenomena in ZnO-varistors can be (semi)quantitatively understood, like
e.g. the high nonlinearities ( >>20), the negative small signal capacitance around UB and the electroluminescence
phenomena observed at the switching point [Gre 1990], [Gre 1998], [Pik 1984], [Pik 1985], [Gre 1984]. The
electroluminescence comes from the fraction of holes, which recombine directly with electrons in the conduction
band, thereby emitting light in the UV-region ( h
3.2 eV ). This direct observation of the band-band
recombination is the most direct evidence for the hole induced breakdown model [Pik 1985], [Gre 1984]. Besides
the valence band states, also defect levels in the band gap of ZnO can be ionized by the hot electrons and their
recombination leads to a strong luminescence in the visible range, like the dominating emission at 700nm (red)
from the doping with Co [Gre 1998], [Pik 1985], [Gre 1984], [Glo 1981], [Cor 1990]. When viewed under the
microscope this strong emission can be observed at every grain boundary along a current filament, which each
light up like a small light emitting diode (see Figure 2.11).
With the model developed by Pike-Greuter-Blatter a good and partly even quantitative understanding is achieved
for the pre-breakdown and breakdown region. This is the most powerful and accepted model today for consistently
explaining the different varistor phenomena. Also the dynamic effects of voltage overshoot under fast pulses can
be understood by this model, although some interface parameters (e.g. interface recombination rates) are not yet
known precisely enough to make the model quantitative for this case [Tua 1988]. No effort has so far been made to
calculate the AC large signal response. Qualitatively the model also explains the slightly higher AC-breakdown
voltage compared to the steady state DC-value UB (see Figure 2.5b) and the observed asymmetries in the resistive
current component, which are due to the charge trapping and de-trapping dynamics [Gre 1990].
Next we briefly address the statistical and microstructural effects, which are caused by the network of grain
boundaries inside a varistor element. In the real microstructure the individual grain boundaries are arranged in a
partly disordered 3-dimensional network and the net current density of a MO-element may well deviate from the
Page 76
local current density seen by a specific ZnO-ZnO grain boundary. Electroluminescence pictures as in Figure 2.11
[Gre 1990], [Gre 1998] made, for the first time, directly visible that the current flow through a MO-block is of
filamentary nature. Similar insight also can be gained by new imaging techniques with high speed infrared
cameras, as shown by Wang et. al. [Wan 1998], Figure 2.13. Several groups [Bar 1996], [Voj 1996], [Che 2002],
[Zha 2005], [Lee 1999], [Wan 1998], [And 2003], [Bog 2000] now have performed simulations of such random
networks in 2-dimensions (e.g. of Voronoi-type), considering different types of irregularities. They provide
interesting insights into various questions, which are difficult to access by experiments, like the role of disorder, the
influence of the grain size distribution, local variations in nonlinearity and switching voltage, fluctuations in barrier
heights, local hot spots etc.
Figure 2.12: Local distribution of current in a random network containing varistor-type grain boundaries
for different positions on the U/I-characteristics: a) ohmic region with a homogeneous distribution of the
current density (V=1.5), b) in the breakdown with clear filamentary conduction (V=3.5), c) higher up in
the breakdown region at the transition to the upturn (V=4.3) and d) in the upturn-region where the
current distribution reverts to a more homogeneous situation (V=6) [Bar 1996]. The gray-level
represents the relative value of the current through a grain boundary, normalized to the total current.
Page 77
Figure 2.13: Infrared thermal images taken with a high speed camera for thin slices of low voltage
varistors with large grain size [Wan 1998]
Depending on the position on the macroscopic U/I-characteristics, the flow pattern through the microstructure can
be quite different and far from homogeneous, as is illustrated by Figure 2.12. In the ohmic regions (U << UB and U
>> UB ) the current flow is rather homogeneous throughout the structure, whereas in the high nonlinear regime
strong current filaments can develop and lead to local overheating, as nicely demonstrated in Figure 2.13. Heat
generation and heat spreading then can become inhomogeneous in the microstructure and may lead to local
thermal runaway and local thermo-mechanical stresses [Wan 1998], [Bog 2000]. For a high irregularity in the grain
size distribution it is possible that two or several filaments join into a single one, which increases the risk of
microscopic hot spots if operated at this condition for a long time (see Figure 2.12). Clusters of several large grains,
as an example, are an obvious condition for creating hot spots [Bar 1996], [Voj 1996], [Che 2002], [Zha 2005], [Lee
1999], [Wan 1998], [And 2003]. For thin samples, as typically used for LV-devices, such statistical disorder effects
are even more critical than for larger volume elements [Wan 1998]. In 3-dimensions some of these reported critical
phenomena are less severe due to the higher number of possible paths available for the current filaments.
However, up to now 3D-simulations have not yet been done for nonlinear random networks, but will certainly come
the more the computational power develops.
In the above network simulations often cases are discussed, where a significant fraction of the boundaries are
assumed to be not varistor-active [Bar 1996], [He 2004], i.e. either ohmic with some assumed conductivity, bad
junctions with poor nonlinearities or even being insulating. The secondary phases like Bi2O3, spinel or pyrochlor
and clusters thereof represent such insulation pockets within the microstructure and this certainly has an influence
on the U/I-characteristics, as experimentally shown by artificially generating such inclusions [Gre 1998]. The reports
of non-active ZnO/ZnO boundaries, however, have to be considered with caution, as experimentally it is extremely
difficult to prepare polished surface structures for local measurements without destroying and short-circuiting the
sensitive grain boundaries. Unfortunately, most authors do not check and comment on such possible artefacts.
Based on the recent understanding that the amorphous grain boundary films are wetting all ZnO-interfaces [Chi
1998], it is rather unlikely that some boundaries should not be varistor-active; local variations in the boundary
properties (barrier heights etc), however are more likely and have to be expected.
Page 78
The simulation studies clearly underline that the optimization and homogenization of the microstructure is
important. This probably is one of the efforts common to all varistor manufacturers. Clear progress has been made
in ceramic process technology over the past decades, as can, for example, be seen in the recent comparative
study on the energy handling capability of major varistor manufacturers [Rin 1997], [Rei 2008] and this report.
Besides the microscopic non-homogeneities discussed above there are also non-homogeneities on the
mesoscopic (ca. 100 m, size of spray granule) and macroscopic ( mm-cm ) scale, which have to be controlled in
varistor manufacturing and which affect the overall performance of a varistor block [Gre 1998], [And 2003], [Ste
2004]. Additional materials challenges to be solved are given by the passivation layer (environmental, dielectric and
thermo-mechanical stresses) and the metallization (adhesion, contacting, rim structure) [Per 2003], [Bog 2000],
[And 2000]. The role of macroscopic (mm-cm) electrical non-homogeneities on the energy handling capability has
been studied by simulations [Eda 1984], [Bar 1996], [Nie 1989] and will be addressed in the following section.
thermal runaway
puncture from current concentrations followed by local thermal runaway and melting
cracking due to localized heating (with or without puncturing)
cracking due to thermo-elastic stresses during high current impulses (even for a perfectly homogeneous
block)
flashover from high dielectric stresses at the rim or surface of the blocks.
An ideally homogeneous block can only fail either by thermal runaway (heat input faster than cooling) or by
fracturing under high current stresses, where the stresses exceed the mechanical strength of the material. In
reality, however non-homogeneities are always present, be it in the microstructure or on the macro scale of the
block. These imperfections will give rise to local overheating followed by either puncturing or/and
compressive/tensile stresses leading to fracturing of the ceramics. Various reports can be found in the literature on
the different types and degrees of non-homogeneities observed and how they translate to mechanical stresses
[Lev 1989], [Eda 1984], [Bar 1996], [Nie 1989]. Figure 2.14 illustrates by IR-measurements the electrical nonhomogeneities, which can be present in a non-perfect varistor block [Ste 2004].
Page 79
Figure 2.14:
Left: Cross section along the axis of a station arrester block ( 63 mm), viewed by a fast IR-camera
upon exposure to a square wave impulse of a few ms duration and a current amplitude of 100-200 A
[Ste 2004].
Right: light intensity profile proportional to the temperature, which in turn is proportional to the current
flowing through the block.
For a complete simulation of the failure behavior, the electrical, thermal and mechanical properties of an
inhomogeneous varistor block have to be calculated in 3 dimensions in a coupled mode. This is a very demanding
task and has not yet been done with finite element methods. Parts of the problem however have been addressed
and are illustrated in Figure 2.15 and Figure 2.16. Figure 2.15 shows the coupled thermal electrical simulation for
the rim of a varistor block, where the metallization is made with an edge margin of 2 mm. The equithermal contour
plots nicely show that this non-metallized rim leads to a current concentration at the metallization edge, where high
local temperatures can occur and can trigger a puncturing.
In general, puncturing has to be expected, if the local heating is faster than the heat spreading and the ceramic
locally is heated to above the melting point (> 750-850 C), where the melt can be progressively ejected, starting
from the surface of the hot spot. The local heating simultaneously also creates thermo-mechanical stresses and, if
these stresses reach the mechanical strength of the material before the hot spot has reached the melting point,
then mechanical fracturing can occur prior to puncturing. Alternatively cracking can follow the puncturing, if
sufficient energy is deposited in the channel. Both phenomena, puncturing and cracking under long duration
impulse, hence have the same origin, but the outcome will depend on such parameters as the local energy input
rate, size and geometry of inhomogeneity, local stresses generated, fracture work needed etc.
A deeper insight in these coupled electrical-thermal-mechanical phenomena was first provided by Bartkowiak et. al.
by using continuum mechanics and assuming simple failure criteria to study the failure modes for the energy
testing of varistor blocks containing hot spots [Bar 1996]. The hot spots were simulated by assuming a small axial,
cylindrical area with reduced breakdown voltage (-5%) in an otherwise homogenous block. Radial, axial and
tangential stresses were then calculated self consistently on the basis of a theoretical U/I-characteristic for different
current densities and two different block sizes. Although the assumed inhomogeneity is rather high compared to
todays high quality blocks (and for symmetry reasons was located in the center of the block), the model predicts
quite well the regimes for the most likely failure modes to occur: thermal runaway, puncture, cracking under tension
and cracking under compression. The results for a distribution arrester disk are illustrated in Fig. 2.16. Note that
depending on the range of current density, different failure modes can show up on a block.
Page 80
Figure 2.15: Equithermal contour plot for an edge margin of 2 mm of a varistor block as obtained from
coupled thermal electrical simulations [And 2000].
Figure 2.16: a) Time to failure t f and b) energy handling capability of a distribution-class type of MO
disk; the different limiting failure modes are shown [Bar 1996].
For the simpler case, where the temperature field across the block is known, the mechanical stresses have been
calculated by Nied [Nie 1989]. Despite the simplifications made the work from Bartkowiak et.al. provides a rather
Page 81
detailed and useful insight into the overall (3D) behavior of distribution and station blocks during the energy
handling capability tests. Note that local hot spots (or channels) can lead to cracking patterns parallel or
perpendicular to the block axis, depending on the impulse amplitude/duration and/or the block size [Bog 2000], [Rin
1997], [Rei 2008], [Eda 1984], [Bar 1996], [Nie 1989].
The above failure modes for energy stresses in the (sub) millisecond to second range have to be clearly separated
from the failures observed under high current impulses, like the 4/10 s pulse. Here a new type of failure mode
appears, which has been identified [Hag 1997] as being a very special case of a thermo elastic stress generation
resulting from the extremely high heating rates produced by this fast impulse: Heating rates up to 10 7 K/s can be
generated inside a block of a distribution arrester! This temperature rise occurs on a time scale much faster than
the material can adapt in its thermal expansion, which is limited by the sound velocity c of the ceramic (c 4-5000
m/s; < h/c, h=block dimension). Even in a perfectly homogenous varistor block this can lead to very high tensile
and compressive stresses. Depending on the block geometry and block properties, these stresses may reach the
material strength and can cause fracture of the varistor block. Hence this thermo-elastic stress is an intrinsic
limitation for all varistor blocks exposed to such short and energetic impulses. This adds the mechanical strength
as an additional parameter for a good material. Existing non-homogeneities in the block not only limit the
mechanical strength of the material, but they can also create additional thermo-mechanical stresses if they lead to
large current non-homogeneities and temperature gradients inside the material [Per 2003], [Bal 2004], [Hag 1997],
[Voj 1997], [Len 2000]. Analytical calculations in simple 1D [Hag 1997], [Voj 1997] or full 3D-analysis [Mah 2001],
as well as 3D finite element simulations [Len 2000], predict these oscillating mechanical stresses in the varistor
body under high current impulse, see Figure 2.17. With the full three dimensional models, the reflections of the
mechanical stress waves from the different surfaces of the varistor body are included, showing that the stress fields
vary between tensile and compressive stresses in time and space in a rather complex manner. Figure 2.17 also
compares the simple 1D-model [Hag 1997], [Voj 1997] with a 3D-simulation, showing that the reflections cause
even higher stresses than predicted from the 1D model only. If the stresses exceed the mechanical strength of the
ceramics, the blocks will crack. Very characteristic cracking patterns can be observed, like the one shown in Figure
2.17, where due to the short duration of the impulse quite often no signs of any discharges are present.
Figure 2.17:left: 1st principle stresses for 1D and 3D model [Len 2000],
right: typical midplane crack under high current impulse for a distribution block with high aspect ratio
(height/diameter) [Hag 1997], [Voj 1997], [Len 2000], [Per 2003].
From the simulations it is evident that the size and aspect ratio of the varistor block have a clear influence on the
thermo-elastic stresses generated. For varistors with an aspect ratio (height/diameter) less than about 0.5, the
st
st
maximum 1 principle stress is normal to the axis, while for a high aspect ratio the maximum 1 principle stress has
axial direction, in agreement with the fracture modes observed (see Figure 2.17). Minimal high current stresses are
predicted for an aspect ratio of ca. 0.9 [Len 2000], [Mah 2001].
Todays simulations do not yet consider possible influences of electrical non-homogeneities, electrodes, metal
spacers, contact forces, encapsulation etc. as present in the real arrester design. Other minor effects also not
Page 82
considered yet in the simulations of the block performance are the slight, beneficial influence of the small positive
temperature coefficient (PTC) in the grain resistivity in the high current range [Cab 2004] or the pressure
dependence of the U/I-characteristics [Dor 1985].
Very little information is available in the literature on flashover phenomena on single MO-blocks [Rei 2008]. Poor
control of the metallization edges of course is one of the possible origins, besides such parameters like contacting
electrodes, ambient conditions, surface passivation, surface contaminations etc. Also care must be taken by
separating a true dielectric surface flashover from discharges being triggered by near edge puncture (as e.g.
expected at high energies for rim situations like in Figure 2.15).
Page 83
Note that most of the degradation or polarization phenomena are reversible and can be healed out by heating to
200-300C without applied voltage. Thereby small thermally stimulated depolarization currents are observed [Lev
1989], either due to the migration of the charged defects or the de-trapping of electrons back to their original
equilibrium configurations. Empirically, some dopants are known for their positive or negative influence on the long
term stability [Lev 1989], [Gup 1990], [Fan 1993], [Bin 1993]. It is assumed that they act either directly via forming
migrating or blocking defects or indirectly via their influence on the defect equilibrium in ZnO (e.g. the density of
Zni) and other phases like Bi 2O3 [Gre 1995], [Gup 1990], [Fan 1993], [Bin 1993]. For a more detailed understanding
of the ageing phenomena, certainly more research work is still needed on such electrochemical processes
occurring near the grain boundaries and the adjacent triple point phases.
1.4
1.2
1.2
0.8
0.8
0.6
0.6
stable
0.4
unstable
0.4
0.2
0.2
0
0
100
200
300
400
500
600
700
800
900
1000
100
200
300
400
500
600
700
800
900
1000
time [h]
time [h]
As mentioned above, in the accelerated ageing tests in the early days of ZnO varistors, the materials showed
increasing leakage currents or power losses vs. time. These changes were found to follow, to a good
approximation, an Arrhenius-type law when varying the test temperature [Lev 1989], [Gup 1990]. Based on these
early observations, accelerated ageing procedures were established in the IEC standard 60099-4. An acceleration
T/10)
factor AF was estimated with AFT = 2.5
, a test temperature of 115C was chosen as well as a test time of
1000h. As an example, the lifetime prediction derived from this test would be equivalent to 110 years at 40C
ambient temperature. In the past, this accelerated ageing procedure provided good confidence on life expectancy
of metal-oxide blocks. For applications with higher ambient temperatures (e.g. 65C) in general the test durations
must be extended to demonstrate acceptable equivalent life times under such condition (see Table 1). A further
increase of the test temperature to shorten the test time seems not applicable, as a change of the ageing
mechanisms at higher temperatures cannot be excluded, making the extrapolation from this high temperature
range to near room temperatures questionable. Also an increase of the test voltage, as another acceleration factor,
is quite limited. With test voltages above the reference voltage, self-heating and thermal runaway may occur and
new transport mechanisms come into play (e.g. hole generation, Figure 2.10).
Page 84
C
40
65
95
h
1000
2000
7000
Table 2.2: Test duration and equivalent time at upper limit of ambient temperature according IEC
60099-4
The problem with the presently used accelerated ageing concept is that most of todays established materials show
a decreasing power loss with time and the decrease normally is higher at higher test temperatures. Hence these
stable materials show behavior just opposite to an Arrhenius-type law, making the standardized test rather
questionable and without a sound physical basis. Presently we do not understand the microscopic origin of the
initially fast and then slower continuous decrease of the power loss observed in most established varistor materials
nor do we have a deeper insight in such cases as shown in Figures 2.18 and 2.19, where the losses go through a
minimum, followed by a steady increase. Such behavior suggests that several mechanisms might be at work on
different time scales. It is well possible that the increase after the minimum corresponds to a mechanism of
instability and perhaps is thermally activated as in the case of an Arrhenius-type mechanism. If this holds then the
standardized test procedure from the early days might still be basically appropriate and only would need some
reformulation. Certainly, further research efforts are also needed here and the existing test procedure should be
kept until an improved understanding puts this sensitive issue on a better founded basis.
High field varistor materials and devices, which allow to build more compact components, like e.g. for
GIS-surge arresters, light weight (gapped) line arresters etc.
ii)
Alternative varistor materials have and will be a topic of ongoing research. Examples, where materials
with reasonably high, controlled and reversible nonlinearities have been demonstrated are varistors based
on SrTiO3, CaCu3TiO12, grain boundary doped SiC, SnO2 etc. Much progress has been made in the area
of SnO2-varistors and electrical characteristics approaching the one of good ZnO-materials have recently
been demonstrated [Bue 2008], [Met 2007]. However, ZnO remains an unique material when it comes to
such practical aspects like availability, costs, purity, particle size, electrical transport properties, economic
sintering temperatures etc and it will be hard to beat it for the big volumes of varistor applications.
iii)
Microvaristors as a by-product of the traditional varistor manufacturing have been developed and are
used as a functional filler in polymers, e.g. for field grading in cable accessories [Str 1995], [Str 2001].
iv)
Multilayer varistors: Impressive and hardly published progress has been made for applications in the
electronics area, where these protection devices of (sub)mm-dimensions are produced on highly
automated production lines in huge numbers or were thick-film varistors with substantially improved
characteristics are evolving.
Page 85
v)
Varistor integration: New concepts and ideas have been evaluated and partly demonstrated for
integrating the surge arrester function into other components [Per 2005]. This becomes possible with the
improvements in the design- and simulation-methods as well as the further progress in arrester
manufacturing. In oil filled distribution transformers integrated solutions show very positive field records in
Japan and USA and proof that integrated solutions can be attractive. Further work, in particular on the test
philosophy, however, is needed to gain a broad acceptance of such new approaches.
Regarding the basic understanding of ZnO varistors, several topics still lack a deeper and quantitative
understanding and hopefully will be addressed in future research activities. A very challenging area certainly is the
ageing mechanism, which seems to be closely linked to the poorly understood defect chemistry in ZnO. The latter
also is of interest for understanding the role of individual dopants and the electrical activation of the grain
boundaries. With the recent discovery of p-type ZnO thin film materials, a revival of ZnO-research has started and
this will also add to a deeper understanding of the atomistic phenomena in varistor materials. A puzzling issue,
which is not conclusively treated in the literature, is the true quality and spread in the junction properties within a
varistor ceramic, where a certain fraction of inactive or bad varistor grain boundaries is reported, however without
quantifying the possible artifacts from sample preparation.
For the transport properties, there is quite a good understanding available today for the DC- and small signal ACbehavior. For the large signal AC-response at 50 or 60 Hz however no attempts have been made so far to apply
the existing junction models to describe the capacitive and nonlinear resistive current components in a quantitative
way. The same holds for the different impulse shapes used in the upturn region, where only a preliminary
simulation study on the transient response has been made so far [Tua 1988]. A related field is the more frequent
use of varistor elements in power electronic circuits, where very steep transients with modest energy stresses, but
very high repetition rates, are typical and only limited knowledge is available today from experiments and from
modeling.
Page 86
2.3.1 FOREWORD
Latest by end of the 1980s MO arresters had definitely been established as state of the art, since their technical
and commercial benefits are quite evident. MO arresters offer low protection levels, high energy absorption capabilities, and stable operation even under severe pollution conditions and lifetimes which easily may exceed thirty
years. Knowledge about principal design of conventional porcelain housed MO arresters can be presumed;
therefore this technology will not be addressed here. However, the very simple structure of an MO arrester the
active part basically consists of a stack of cylindrical MO resistor elements supported the development of polymer
housed arresters at a very early stage. They were introduced for the first time in the mid-1980s in the distribution
voltage level. After about 15 years of development some few basic design principles of polymer housed arresters
can be distinguished with a variety of individual sub-solutions. While there are less basic design variants for
distribution than for HV arresters (only cost effective, no "high tech" solutions are applied), the variety of subsolutions is much larger, one reason for this given by the fact that there are far more manufacturers of distribution
arresters on the market than in HV. Accordingly, more design variants had to be created in order to avoid patent
conflicts. As mentioned, for HV arresters at least one more basic design principle is applied the use of composite
hollow core insulators but there are not so many sub-solutions, as there are not too many manufacturers of HV or
EHV polymer housed arresters worldwide. Therefore a classification of today's polymer housed HV arrester
designs is comparatively easy. It must be mentioned, though, that there still does not exist any official
nomenclature. Designations like "Type A" or "Type B", as they are used in this contribution, must not be mixed up
with other emerging classifications. For instance, the new IEC document 37/317/CDV on arrester short-circuit
testing [IEC 37] has introduced "Design A" and "Design B" arresters. These designations have different meanings,
as they serve for classification with regard to short-circuit performance only.
This contribution exclusively focuses on constructional design principles. Other important design aspects such as
the different performance characteristics can be looked up in [Hin 2003].
Page 87
help of compression springs. Additional supporting elements (not shown in the figure) may be necessary to fix the
insert in radial direction. What is important and sometimes criticized is the fact that this arrester due to its enclosed
gas volume needs a sealing and pressure relief system. This has in fact to be designed and manufactured with the
same care as it is the case with porcelain housed arresters. But just for this reason this does not constitute a real
problem for an experienced manufacturer of HV arresters. There are numerous makes of HV arresters on the
market, which have an excellent service record (over twenty years or even more) also with respect to their sealing
systems. Therefore, whether this arrester design has problems with moisture ingress or not is a matter of the
manufacturer's know-how and production quality, as it has always been with porcelain housed arresters and by
the way as it is also with the other designs of polymer housed arresters, which by far are not all inherently leak
age sealed.
Porcelain;
Polymer Type A
Polymer Type B2
MO column
Gas
Solid/semi-solid material
Outer housing
Type A: gas volume enclosed, separate sealing system, pressure relief vents
Type B: no intentional gas volume included
...1: wrapped mechanical structure or tube
... a: FRP material directly wrapped onto the MO blocks
... b: FRP tube with distance to the MO blocks, gap filled by other material
... 2: cage design
Figure 2.20: Classification of basic arrester designs; cross sectional side and top views
Page 88
MO resistor column
Composite
hollow core insulator
(FRP tube/ rubber sheds)
Page 89
MO column
FRP wrap
Page 90
The resulting gap between MO and FRP material is then filled by solid or semi-solid material. Again, slots may be
provided on the surface in order to improve short-circuit performance.
Type B2: this is a completely different design concept and usually called the "cage design". While the mechanical
strength for Type A and Type B1 arresters is exclusively provided by the FRP structure (closed or partly open tube)
this is done by the MO resistors themselves with Type B2. For this purpose, they are clamped between the metal
end fittings by FRP loops or rods, applying an enormous axial pre-stress in the range of 100 kN [Ste 2003]. The
basic mechanical design is shown in Figure 2.23. The left variant is designed with loops, which are fixed in notches
in the end flanges. This design was first introduced for distribution arresters [Sch 1996] and then extended to HV. A
sub-variant (Figure 2.23, middle) uses an additional bondage from polymeric material in order to achieve the
mechanical and short-circuit characteristics required for application in HV and EHV [Sky 2002]. Also the other
variant of Type B2 arresters (Figure 2.23, right) was first realized in distribution arresters and then further
developed for HV. Here, FRP rods are applied, which are mechanically fixed in holes in the end flanges by a
proprietary clamping system. Figure 2.24 shows photographs of real arrester modules as they are produced by two
different manufacturers. Main technical advantages of the cage designs are that they offer comparatively high
mechanical strength combined with an inherently good short-circuit performance.
Sealing
Bonded
by
molding
process
Gap filled
by solid
material
With regard to the commercial aspects it is impossible to give any statement here on production cost, which greatly
depends on the total manufactured quantities, the degree of automation and process optimization, the quality of the
applied materials, the degree of type diversification and so on. However, it can be noticed that the Type B arrester
in general constitutes the most economical way to produce an arrester. At the same time it offers a technical
performance, which in most cases ranks between comparable porcelain and polymer housed Type A arresters. The
Type B arrester is therefore the typical "low cost" arrester mentioned earlier, one of the reasons for the success of
this design on the market.
Page 91
Page 92
The differences in the design differentiation of HV and MV arresters are due to the fact that in HV the mechanical
strength of the design is of more importance, while in the MV field the production process is of higher importance
(for cost reasons). The design principle B2 for HV arresters is almost identical to Group I for MV arresters.
2.3.6 CONCLUSION
Polymer housed HV arresters actually have a market share of roughly 30%. First installations have been made
around 1990, and so far there is no indication that they will not show the same good performance as MO arresters
in general or polymer housed distribution arresters, which have successfully been in service for nearly 20 years
now. The technology can be considered mature. Thus it can be predicted that the share of polymer housed
arresters in HV will continuously increase, because they usually offer economical as well as technical advantages
over porcelain housed designs.
Different design principles have emerged in the meantime, which basically can be divided in Type A arresters
("tube design"), Type B1 arresters ("wrapped design") and Type B2 arresters ("cage design"). The Type B designs
usually have lower and the Type A designs higher market prices than comparable porcelain housed arresters.
These different designs normally serve different market segments, which can be classified as "low cost" or "price
orientated" on one hand and "high performance" on the other. Primarily this is a matter of the system voltage level.
In systems of 72.5 kV Us 300 kV ("HV" systems) mostly the "low cost" variant is preferred, since this is the
domain of standard applications where no exceptional requirements on electrical or mechanical performance exist.
The majority of polymer housed arresters have been optimized for this market. The arrester designs coming into
question here are the Type B designs, where in general the B2 variant offers higher mechanical strength.
In systems of Us > 300 kV ("EHV" systems) mechanical requirements usually favor the Type B2 or even the Type A
designs. Especially with the latter nearly any required mechanical, electrical or safety feature can be achieved,
which has its price, however.
Page 93
2.4.1.1General Comments
These two distribution-type arrester are similar in that each is typically used to protect connections between
underground cables and ground-mounted (pad-mounted) or underground transformers. As such they are usually
directly attached to the cable connectors on each phase and between high potential and earth. In this sense they
offer optimal protection of the cables and the transformers since the protection distance (between arrester and
equipment) is very small. It is worth noting that the installation enclosures can easily become submersed during
flooding or other events.
Appropriate standards for separable connectors, of which these arresters form a sub-group, are IEC 60502-4, IEC
61442 and IEE 386. These contain related test procedures. Appropriate arrester standards are IEC 60099-4 and
IEEE 62.11. They contain definitions for separable and/or dead front arresters as follows:
-
IEC an arrester assembled in an insulated or screened housing providing system insulation, intended to
be installed in an enclosure for the protection of distribution equipment and systems. Electrical connection
may be made by sliding contact or by bolted devices; however, all separable arresters are dead-break
arresters.
IEEE an arrester assembled in a shielded housing providing system insulation and conductive ground
shield, intended to be installed in an enclosure for the protection of underground and pad-mounted
distribution equipment and circuits.
Note: these arrester types are often referred to as elbow arresters, with reference to their characteristic form or
shape.
Figure 2.27: Appearance and internal design of a typical dead front, separable metal-oxide arrester. In
this example the connector is also equipped with load-breaking capability.
Page 94
Various methods are used to contain the MOV disks within the housing. They may be contained by composite
elements similar to those used in overhead distribution arresters. Alternatively they may be directly bonded to each
other with appropriate conducting adhesives. Less favorably they may only be contained within and by the housing
itself and its ground termination elements.
The major difference between these two arrester types lies in the presence or not of a screened or earthed shield
that provides solid protection whilst deliberately or accidentally making contact with the device whilst under
potential. The external screen is either a discrete element separately molded onto the insulated housing or simply
but less favorably painted onto this housing.
Page 95
For this reason, in cases where the fault currents can be high i.e. in low-impedance earthed systems, the failure
mode during short circuit testing is required to be benign with respect to the immediate surroundings. This usually
means that the housing of the device may either not vent any debris at all or it may only do so in a downward
direction away from the neighboring phases.
HCSD test
The conditions of this test are very similar to those for other arrester designs (apart from the ambient temperature).
However it is worth noting that generally high current impulses do to penetrate underground systems to the same
extent as overhead systems. The rated high-current short-duration impulses should therefore be lower. For
example, 40kA 4/10 s may be considered adequate where an overhead system requires 65kA 4/10 s.
For dead front-installations this test is more severe than for live-front. The reason is that the proximity of the shield
or ground plane increases the electrical stresses on the outer edges of the MOV disk to much higher values that
can cause it to fail, where it would otherwise withstand the same impulse current conditions in a live-front
assembly. This means that the dead front arrester must be HCSD tested with a sample (model) representing the
same electrical situation as in the final
Figure 2.28: Appearance and internal design of a typical under-oil metal-oxide arrester.
The under-oil arrester has been employed since 1980 in the USA, with more than 500000 units installed, and since
1987 in Japan with an even larger quantity. They have in each case demonstrated excellent reliability as devices
Page 96
and have significantly reduced the lighting-related damage to transformers. In the case of Japan, this has been
impressively documented [Ish 2004].
Page 97
Figure 2.29: Typical installation of a GIS arrester at the line entrance of a gas insulated substation (Us =
420 kV), example Siemens.
For GIS arresters a main engineering target is a compact, space saving design [Sch 1992]. GIS arresters
principally consist, as all other MO arrester designs, of one or more parallel columns of MO resistors installed within
a housing, in this case an earthed metal vessel filled up with SF6 gas. The individual columns are built up by
connecting MO resistor elements in series. In order to achieve an economical and space saving design and to
minimize the impact of stray capacitances to the earthed vessel often a meandering mechanical design of the
active part has to be used. This reduces the overall physical length of the active part and additionally contributes to
1
reduced self-inductance . Insulating plates of extreme high electric withstand have to be applied in order to insulate
the layers of MO resistors from each other. Figure 2.30 shows the principle design of the active part of an EHV GIS
arrester. Each of the insulating plates is electrically stressed by the voltage drop across eight MO resistors.
The self-inductance per unit length of GIS arresters is typically assumed as 0.3 H/m, while air insulated designs have a
typical value of 1 H/m.
Page 98
insulating plates
MO resistors
current path
Figure 2.30: Design of the active part (mechanically three columns, electrically one column) of an EHV
GIS arrester using conventional MO resistors.
Due to the very short radial distance between the active part on high voltage potential and the earthed vessel a
relatively high capacitive stray current is flowing, which leads to an unfavorable axial voltage distribution along the
active column of MO resistors. For this reason countermeasures such as metallic grading elements (hoods or rings)
or capacitive grading elements have to be taken. Figure 2.31 shows the principle design of GIS arresters for
different system voltages if standard MO resistors with 200 V/mm are used. For systems up to 170 kV system
voltage GIS exist with only one phase or with all three phases in one metallic enclosure. Accordingly, the GIS
arresters are designed in the same way. Up to a system voltage of Us = 170 kV the MO column consists typically of
a linear column. For higher system voltages generally a mechanically three or four column design is used [Gh
2006].
insulator
electrical connection,
high voltage
pressure relief device
grading hood
MO column
vessel
ground plate
Figure 2.31: Design of GIS arresters (principle) for different system voltages. Left: single phase design for
system voltage up to 170 kV. Middle: Three phase design up to 170 kV. Right: single phase design for
system voltages above 170 kV with electrical one phase, but mechanically three columns of MO
resistors, courtesy Siemens.
Page 99
Actually, there are two major directions of MO resistor development: increasing the energy handling capability (in
terms of kilojoules per cubic centimeter of volume) and increasing the field strength (in terms of volts per millimeter
of height). Typically, a MO resistor for high-voltage arrester applications has a field strength of 200 V/mm at a direct
current of 1 mA. But MO resistors of 400 V/mm have successfully been developed in the mid-1990s [Ima 1984] [Shi
1997] and are commercially available, and MO resistors with 600 V/mm are reported [Fuk 2012].
The benefit of this progress can less be utilized for air insulated arresters as the high field strength causes severe
dielectric stress across the external surface and the heat after energy injection cannot be dissipated to the ambient.
But GIS arresters, where dielectric problems along the active part do not occur due to the high electric strength of
the surrounding SF6, and where the heat transfer is much better than in air, take advantage from the fact that the
overall length of the MO column can be drastically reduced when using high field MO resistors. With the new high
field MO resistors GIS arresters for application in voltage systems up to 550 kV can be built using a simple linear
stack of MO resistors instead of the meandering mechanical design.
Figure 2.32 shows the difference in dimensions of a GIS arrester containing an active part built up with MO
resistors with normal field strength (200 V/mm) and three column meandering design (left), and designs using
high field MO resistors of 400 V/mm (middle) 600 V/mm (left). All of these designs are for application in 550 kV
systems and have the same energy rating. Besides the simpler and space saving design the SF6 volume is
reduced drastically, which is an important argument in the todays discussion about greenhouse gasses.
Figure 2.32: GIS arresters for 550 kV systems with MO resistors with normal field strength (left) and
high field MO resistors of 400 V/mm (middle) and 600 V/mm (right), courtesy Toshiba.
GIS arresters with meandering mechanical design using high field MO resistors are designed to reduce the arrester
height and increase the mechanical strength against seismic stresses even in horizontal installations, which gives
more flexibility in positioning of the arresters in an optimized GIS layout, see Figure 2.33.
Page 100
Figure 2.33: GIS arresters for 550 kV systems with MO resistors with normal field strength of 200
V/mm and 400 V/mm in a mechanical meandering design, courtesy Mitsubishi.
Page 101
There are various potential attractions of these devices; for example reduced cost or size, both increasingly more
important requirements. Improved reliability, improved overvoltage protection, improved environmental protection
have all been claimed as well. However, most likely the major benefits have still to be demonstrated, since the
increasing freedom of the above developments afford the engineer more opportunity for innovative design.
However, integrated arrester systems bring special concerns with them; not the least of which is how to effectively
test a device with multiple, sometimes interrelated functionality. These and other special considerations have been
reviewed in a recent publication of this Working Group [Per 2005].
Perhaps the most obvious of dual applications for MO surge arresters is the use as post or suspension insulators,
given their high cantilever strength and dielectric withstand. In Figure 2.34 and Figure 2.35 examples are given.
Figure 2.34: MO arresters used as post insulators in a 420 kV substation (left, example Siemens) and as
suspension insulator/line arrester in a medium voltage trial line in Norway (right, example ABB).
Page 102
Figure 2.35: MO surge arresters with porcelain housing and modified grading ring integrated in a center
break disconnector. System voltage 420 kV, example Siemens.
Arresters have a limited spatial protection range due to travelling wave effects. A lightning impulse voltage may
reach twice the value of the related arresters lightning impulse protection level at the terminals of the device to be
protected, depending only on the steepness of the incoming overvoltage and on the distance between arrester and
the device. This protective zone or separation distance is typically in the range of several ten meters in high
voltage applications down to only a few meters in distribution systems. Therefore, the integration of MO surge
arresters directly into other equipment of the substation improves naturally the protection of the substation. This is
of course an additional very important benefit besides the space saving for the substation.
Page 103
3.
3.1 Summery
This part of the Brochure covers energy handling capability of MO resistors and arresters. It starts with an
introduction and a short subsumption of the different aspects of energy handling, basically divided into "thermal"
and "impulse" energy stress. It then reviews the state of knowledge by evaluating some of the most important
published literature on this subject. Cigr WG A3.17 has initiated an experimental research program on energy
handling, which is being performed at Technische Universitt Darmstadt. It is the most comprehensive investigation
on this subject performed so far. Though this project is still going on, some important results can just be
summarized. Several thousand MO resistors from different manufacturers worldwide were tested. The test
specimens were of approximately 60 mm diameter and 45 mm height, as typically applied in HV arresters, and of
approximately 40 mm diameter and 45 m height, as used in high duty distribution arresters. Energy was injected by
long duration and double exponential current impulses, among them also the new "lightning discharge" impulse
(sine half-wave of approximately 230 s base time, resulting in an impulse current 90/200 s) that was recently
introduced to the standard IEC 60099-4 (Annex N). Furthermore, alternating current stress was imposed as well, in
order to investigate the impact of this kind of stress, but also to check if this test approach can be favorably applied
in the future. A very important aspect was the introduction of a "complex failure criterion", which means that the MO
resistors were not only stressed up to visible mechanical damage, but that also deterioration of the electrical
characteristics was considered. The investigations have basically confirmed the typical dependence of energy
handling capability from current density, as published before. But there are exceptions at extremely high current
densities, where in many cases the coating of the resistors would fail. A general increase in energy handling
capability, expressed in terms of 50 % failure energy, by 20 % in average and up to 70 % in some cases, compared
with values published in the late 1990s, can be observed. The "complex failure criterion", however, leads to more
pessimistic statements, as it turns out that the MO resistors for distribution applications typically (but with
exceptions) show remarkable degradation of their electrical characteristics before they fail mechanically. Finally,
this chapter ends with a critical review of the existing arrester standards with regard to energy handling definitions
and test procedures. Some lacks are identified and suggestions for improvement are given. Based on the actual
and updated knowledge of energy handling capability it should be possible to improve the standards in their next
revisions accordingly in order to better fulfill the requirements and expectations from manufacturers and users of
MO surge arresters.
3.2 Introduction
Since simple spark gaps for overvoltage protection were replaced by surge arresters, the arresters' energy
handling capability has become an important issue. While in series gapped SiC-arresters, especially in the EHV
systems, the energy during charge transfer to ground is shared among the arcs burning in the gaps and the series
connected SiC-resistors, this energy has to be dissipated exclusively by the MO-resistors in case of gapless MOarresters. On one hand, this results in high requirements on the non-linear resistors, which have to act as nearly
perfect "insulators" under normal operating conditions and as high-performance overvoltage limiting "energy sinks"
under overvoltage stress. On the other hand, one should expect that energy handling definitions, specifications and
test procedures would have become simpler, as only one element the MO-resistor has to be considered. When
looking to the published literature and to the actual surge arrester standards, one will find that this is obviously not
the case. There is still a certain lack of general knowledge and theoretical understanding about some energy
handling capability aspects, for instance the impact of the way the energy is injected or degradation effects caused
by multiple or repeated energy stress. Of course, the theoretical background has been improved, simulation tools
have been developed which allow many effects to be modeled and simulated, and finally thirty years of experience
with MO arrester application have given a high degree of confidence in their reliable performance. However, the
fact should not be underestimated that MO resistor manufacturing requires a rather complex technology, and
therefore the final products' performance will always strongly depend on production technology and quality. With
respect to the MO arrester standards: they had to be developed in a time when MO technology was quite new and
still emerging, and it took about ten years after introducing the first MO arresters to the systems that first related
2
standards were published. Till now, the standards do reflect this situation. In terms of energy handling issues, one
still recognizes historical approaches from the gapped SiC arrester era. It is, therefore, the time to think about new
definitions of energy handling capability and about revised, appropriate test procedures, based on actual
knowledge and most recent findings. Besides others, this was the task given to Cigr WG A3.17: to check the
actual literature and standards about energy handling issues, to contribute to some of the open questions by a
comprehensive practical research program, and to work on proposals for energy handling issues in future revisions
of the international arrester standards. The basic results of this work are reported below.
Page 105
a factor of ten to twenty at 150 C compared with 20 C. This power loss characteristic is specific to a particular MO
material and make.
Under continuous operating conditions, an arrester will adopt an operating temperature slightly above ambient
temperature, and the generated heat can easily be dissipated to the ambiance: the arrester adjusts itself to a stable
operating point (left intersection of the two curves in Figure 3.1). However, once a high amount of energy is injected
into the arrester under overvoltage conditions, the arrester temperature will be increased in form of a step function,
with typical values of temperature increase under nominal energy stress up to 100 K or even more. The operating
point will instantaneously jump to the right on the electrical power loss curve. As long as it remains left of the
second intersection point of the two curves, the generated heat can still be dissipated to the ambiance, and the
arrester will cool back to its normal operating temperature within five of its thermal time constants. But if the right
intersection point the limit of thermal stability is reached or even exceeded, the arrester will generate more heat
than can be dissipated and electrical power losses will further increase and finally destroy the MO material by
excessive heat (puncture at several hundred Degrees Celsius).
It is evident that, on one hand, the thermal stability limit depends on the overall arrester design. Arresters with MO
resistors directly covered by a polymeric housing, for instance, will have a thermal stability limit at higher
temperatures than conventional porcelain housed arresters, since they can better transfer heat from the MO
resistors to ambiance. On the other hand, also the MO material properties (electrical power losses and their
temperature dependence) have an effect, because the more pronounced the increase of power losses with
temperature is, the more will the right intersection point of the two curves be shifted to the left, i.e. to lower
temperatures. As well, the curve of electrical power losses versus temperature is affected by possible impulse
degradation, i.e. it will be shifted upwards [Hei 2001], which again changes the limit of thermal stability to lower
temperatures.
However, definition and verification of the thermal energy handling capability is a comparatively easy task. Injected
energy per volume and temperature increase are simply linked by the heat capacitance, which has a non-linear
dependence of temperature, and can, acc. to [Lat 1983], be calculated as
W
V
2,59
J
J
0, 0044
cm K
cm K
(equation 3.1)
Energy
per volume in J/cm
Energie/Volumen in J/cm
900
800
700
600
500
400
300
200
100
0
0
10
20
30
40
50
60
70
80
90 100 110 120 130 140 150 160 170 180 190 200 210 220 230
Temperature
Temperaturin
in C
C
Figure 3.2: MO resistor energy per volume vs. temperature, acc. to [Lat 1983]
In order to verify thermal energy handling capability, energy may thus be injected into the arrester by any suited
method that will rise its temperature to a value related to the specified energy, because the only purpose of this
Page 106
verification is to demonstrate that the arrester is able to cool back afterwards. Of course, the operating conditions
(applied power-frequency voltage) must be specified, and possible electrical aging of the MO material, e.g. by
current impulse stress, must be considered by appropriate conditioning procedures.
Figure 3.3: Examples of MO resistors, mechanically failed by single impulse energy overload. Left: failed
by thermo-mechanical cracking; Right: failed by flashover of the coating.
When looking deeper into the details, however, the matter is more complex. One issue is the definition of a
"failure". Not in all cases will the MO resistors fail so obviously as to be seen in Figure 3.3. There may be only
some small punctures in or at the edge of the metallization, but furthermore the MO resistor may look intact. In
other cases, no damage at all might be seen by a visual examination, but the MO resistor is pre-damaged and will
not pass any further energy input. Or its electrical characteristic may be dramatically changed such that if this
happened in a complete arrester the arrester would become thermally instable.
At this point it shall be noted that in general use of any impulse energy handling capability higher than the "thermal"
energy limit can be made only if the arrester is not applied to an operating voltage close to its continuous operating
voltage, since otherwise the arrester would suffer a thermal runaway even if its MO resistors were able to handle
the excessive impulse energy input.
Another point that has to be addressed is if "withstand" capability shall be specified in a deterministic way
meaning that no single failure is allowed when the MO resistor is stressed by its withstand energy or if a statistical
approach is more appropriate, in the same way as for the dielectric strength definition of external insulation (in
which case "withstand" voltage stands for a "10 % flashover probability" voltage). Verification of a withstand energy
is a difficult task anyway, as even for the statistical approach acceptable failure rates of individual MO resistors in a
complete arrester at "rated" energy handling capability are in the range of only 0,1 % or less. This shall be
demonstrated by the following example. An arrester for a 420 kV system is made up from approximately n = 65 MO
resistors. If each resistor has a failure probability of 0,1 % (p = 0,001) at its "rated" energy, the full arrester, at the
n
same rating, will have a failure probability of P = 1 (1 p) = 0,063 or roughly 6 %, respectively. Higher failure
probabilities for the full arrester are hardly acceptable! Vice versa, if the full arrester shall have a failure probability
of only 1 % (P = 0,01) at its "rated" energy the individual MO resistors in this case must have a failure probability
1/n
-3
p = 1 (1 P) = 0,15510 or approximately 0,015 % only.
But so far, there is no effective test procedure to reliably verify failure probabilities of only 0,1 % or even less.
The observation that the actual failure rate of high-voltage arresters in service is obviously close to zero can be
explained by one or more of the following reasons. In general, the energy stress in real service may be far below
Page 107
the actual impulse energy handling limits that are verified during type tests. Energy of the "withstand" level may be
injected only few times during the arrester's life time. The actual failure probability at "rated" energy may be much
lower than can be verified by the type tests. The type tests as actually specified therefore obviously ensure good
operational performance in service, but they may cause overdesign of arresters, and they will not give any
information about the real limits of energy handling capability.
Furthermore, the question comes up how helpful to the user the information is that an arrester passes one single
energy input, but not a second one a certain time later on. This might be interesting in only some few special
cases, but in general, the expectation will be that an arrester can be stressed by its "withstand" energy several
times during its total service life. But what is the meaning of "several": three times, ten times, eighteen times (the
actual number of energy stresses in the long duration current impulse withstand test according to [IEC 2009]), or
even, e.g., one hundred times? It is assumed to date that energy handling capability decreases with the number of
stresses, but in an actually unknown dependence (this is being investigated in a follow up research program of
Cigr WG A3.25 and will be published later). It is also questionable how meaningful a type test on three MO
resistors by eighteen energy injections each is (this is again the long duration current impulse withstand test
according to [IEC 2009]), as the impulse withstand capability is not only a material issue but depends at least to the
same degree on production quality.
Finally, multiple stress (i.e. energy injections in time intervals of only a few milliseconds, as it may happen by
multiple lightning strikes) has turned out to result in interesting effects, which are not only related to the MO
material properties but to the overall system of the MO resistor and the arrester [Dar 1998].
Thus, definition of impulse energy handling capability is by far not trivial, and the same is true for appropriate test
procedures. Users have become familiar, for instance, with the line discharge classes of [IEC 2009]. These are
very helpful and easy to apply in standard applications. But more and more users have very special system
configurations, and in many cases, by means of system analyses they are in a position to give detailed information
on the energy duties from the system. However, none of the to dates arrester standards gives satisfying answers
and information on the various different aspects of energy handling, and it must be the objective of any research on
MO arrester energy handling capability to give better guidance on this matter in the near future.
In a comparison between SiC and ZnO arresters, Sakshaug [Sak 1989] concluded that ZnO resistors in general
have a higher energy handling capability than SiC resistors. He gave a value of (170200) J/cm for the thermal
capability and mentioned at the same time that under alternating current stress up to mechanical failure values of
(450700) J/cm were observed.
Eda [Eda 1984] performed one of the very first methodical experimental investigations about energy handling
capability of MO resistors. For this purpose, he produced resistors of (10110) mm in diameter and of (120) mm
height, i.e. he worked with non-commercial test specimens. He reported about two different failure modes under
impulse energy stress, that is to say cracking and puncture. Flashover as a possible failure mechanism was not
observed on these specimens. Most of his published results are related to small discs of only 1,3 mm height. He
found an impulse energy limit value of 750 J/cm for discs of 1,1 cm diameter. For discs of 2,76 cm diameter, he
found an energy limit of 520 J/cm when stressed by a 2 ms impulse and of 615 J/cm for a 20 s impulse. These
findings indicate two important tendencies: energy handling capability decreases with increasing diameter and
volume, respectively, which can be explained by a worse homogeneity of the ceramic material with larger diameter
and volume, and it increases with shorter impulse durations or, in other words, with higher current densities. Any
3
All MO resistors are basically made from ZnO. In the beginning, "ZnO" arrester and resistor was a common terminology. To
date, one usually speaks exclusively of "MO" arresters and resistors in this context.
Page 108
further interpretation of the results and comparison with modern MO resistors seems problematic as Eda used very
thin discs where the contact system assumedly had strong influence on the experimental results (e.g. due to a
comparatively high amount of heat transfer). Furthermore, it is not for sure that self-made resistors of 114 mm
diameter and 10 mm height can really be compared with to dates commercial products of the same dimensions.
Studies of Martinez and Zanetta jr. [Mar 1996] basically confirmed Eda's findings.
One of the best known systematic investigations was performed by Ringler et al. [Rin 1997]. Ringler's group
investigated 350 commercially available MO resistors from three different manufacturers. Their diameters were in
the range of (6264) mm, and their height was (2324) mm. The rated voltage was approximately 3 kV. Energy
stress tests until mechanical failure were performed with very long duration current impulses and with 60 Hz
alternating current on a batch of (25...50) specimens each at current amplitudes of 0,84 A up to 35 kA,
corresponding to current densities of (0,031130) A/cm.
Failure energies varied from approximately 460 J/cm at a current density of 0,03 A/cm up to nearly 1700 J/cm at
a current density of 1130 A/cm. There was a distinct increase of energy handling capability with increasing current
density and thus decreasing stress time duration. These findings basically correlate with those reported by Eda in
1984.
At this point it must be recalled that failure energies (e.g. related to 50 % failure probability) are of course far
beyond those energies that are the basis for an arrester specification. The latter are in the range of only 200 J/cm,
for reasons that were explained above. One finding of Ringler et al. by statistical evaluation was that varistors with
a 50 % failure probability of more than 400 J/cm may have a failure probability of still 1 % at 200 J/cm.
When the average current amplitude during energy stress is plotted as a function of the average time to failure in a
double logarithmic scale, this gives a linear dependence in form of a sloping straight line over five orders of
magnitude. This relationship (log I = const. log t) was published by Ringler et al. for the very first time and since
then belongs to the basic knowledge about energy handling capability of MO resistors. The failure charge for
Ringler's investigations was (1117) As, corresponding to failure charge densities of (0,350,55) As/cm.
Ringler et al. reported furthermore that under alternating current stress the resistors would mainly fail by puncture
close to the edges. Under impulse current stress up to 35 kA they reported about little holes on the surface of the
metallization, including the edges. This may have been caused by the quality of the metallization edge. It is well
known today that energy handling capability can be increased by optimization of the metallization. In most cases
the investigated resistors failed by puncture or by some kind of "tracking" along the outer coating. In only few cases
mechanical cracking was reported to be the failure mechanism.
Another important observation of Ringler et al. was that none of the considered distribution functions Normal,
Weibull and Gumbel could be given a preference. Failure probabilities of MO resistors cannot consistently be
described by any of these distributions. All these distribution functions covered the observations with deviations in
different details, but the final outcome (e.g. in terms of 50 % failure energy) was comparable for all of them.
Therefore, it was suggested to apply the Normal distribution as a generally known function and acceptable
approximation when results about failure energies shall be compared.
Boggs et al. [Bog 2000] investigated energy handling capability depending on how the metallization is
implemented. It was the objective of these investigations to find out if the metallization should preferably be applied
exactly up to the edge of the MO resistor or if a certain clearance to the edge should be kept, and how this would
affect energy handling capability. Performed simulations indicated a distinct temperature increase directly at the
edge of the metallization. It was finally proposed to keep a distance of (0,30,6) mm to the edge of the MO resistor
in order to get an optimized energy handling capability. For smaller distances, it was stated that dielectric strength
would be affected (risk of external flashovers). This conclusion has not generally been accepted. Many MO resistor
manufacturers have successfully implemented a metallization exactly up to the edges. This has to be seen not only
from the point of optimized energy handling capability; it is also a concern of manufacturing technology, since
implementation of a metallization up to the edge is difficult. However, Boggs' investigations have shown that the
way of metallization does have an influence on energy handling capability. If the metallization ends too far away
from the MO resistor's edge local current densities may reach values that increase the risk of puncture of the MO
Page 109
material. Also the quality (smoothness) of the edge on a microscopic scale has effect on the overall energy
handling capability.
Bartkowiak et al. [Bar 1996a] worked intensely on the simulation of energy stress and energy handling limits. Two
different kinds of MO resistors were considered: one with a diameter of 32 mm and a height of 45 mm,
representative for medium-voltage (distribution) arresters, and the other with a diameter of 63 mm and a height of
23 mm, typically applied in high-voltage (station) arresters. Apart from the different typical applications, these MO
resistors thus differ distinctly in their aspect ratios, i.e. their ratios of height over diameter, which was considered in
particular and which has influence on the failure mode as could be demonstrated by Bartkowiak at al.
With regard to failure mechanisms, it was distinguished between puncture, cracking and thermal instability (where it
has to be noted that thermal instability is not an impulse energy failure mechanism as such; it just starts a process
that finally leads to puncture of the MO resistor by overheating). Only heat dissipation in radial direction was
considered, as it is the case in a real arrester (apart from its ends). Material inhomogeneity was modeled by a
straight small channel in the center of the MO resistor, having a varistor voltage reduced by 5 % compared with the
overall characteristic. Energy injection was simulated by a direct current, heating the material. A further boundary
condition was the possibility of free movement of the material in any direction. It must be critically noted that in real
arresters this is not always the case. In many designs the MO columns are mechanically clamped by extreme
forces in the range of 100 kN or even more. Therefore, in a real arrester any thermal expansion of the MO material
may be limited or even totally suppressed, which may result in different distributions of pressure and tensile forces
in the material. This must be kept in mind when interpreting the simulation results.
Some more details about the simulation are explained in [Bar 1999]. Following preconditions were assumed to
result in failure of the MO resistor:
-
a tensile force larger than 480106 N/m in axial direction will cause cracking,
a tensile force larger than 140106 N/m in radial direction will cause cracking,
an average overall temperature above 190 C will cause thermal instability,
a temperature of the center channel above 800 C will cause puncture.
The performed simulations showed that the minima of energy handling capability depend on the failure mode. For
the distribution MO resistors, the minimum was found to be 310 J/cm at a current density of approximately
1 A/cm. The related failure mode is "cracking". Energy handling capability then increases with current density. For
"puncture", the minimum failure energy is found at a current density of 1 A/cm as well, but at a higher level of
about 600 J/cm. For the failure mode "thermal instability" the failure energy is about 980 J/cm at extremely low
current densities of less than 0,0001 A/cm, and it then reaches a nearly constant value of 580 J/cm over the full
range of current density from 0,001 A/cm up to 50 kA/cm. For distribution MO resistors the minimum of energy
handling capability thus will be found at current densities of about 1 A/cm, for puncture as well as for cracking.
This is not the case for the high-voltage resistors. For the failure mode "puncture" the minimum was calculated to
be 420 J/cm at a current density of about 0,1 A/cm, while for "cracking" the minimum energy is slightly higher
500 J/cm but at much higher current densities of about 20 A/cm. These findings, however, could not all be
verified by recent experimental investigations, as will be reported later in section 3.4.2.
Bartkowiak also modeled the behavior of an MO resistor on the basis of a two-dimensional, randomly generated
4
Voronoi network [Bar 1996b] [Bar 1006c]. The network is made up from three components: "good" grain
boundaries with extremely non-linear voltage-current-characteristic, "bad" grain boundaries with poor non-linearity
and "ohmic" grain boundaries.
An interesting recent publication is from China [He 2007]. He and Hu report about tests on two different makes of
commercial varistors: type A with a height of 10 mm and 32 mm in diameter; type B with a height of 10 mm but a
diameter of 52 mm. For energy tests with long duration current impulses of 2 ms and 8 ms time duration they quote
cracking and puncture as dominating failure mechanisms. For the MO resistors of 32 mm diameter they give
surprisingly low failure energy values of only (216575) J/cm. Such low values for these comparatively small
4
Voronoi polygons acc. to the Russian mathematician Georgi Feodosjewitsch Woronoi (1868-1908). Voronoi polygons are
applied in material sciences to simulate a random crystal constellation in polycrystalline materials.
Page 110
resistor elements are interesting, as they show that though MO resistor technology is considered to be mature, one
has always to be aware that products will come to the market that do not fulfill the general expectations on energy
handling capability. It is thus once more important to have clear definitions and related test procedures in future
arrester standards that allow an easy and simple evaluation of energy handling capability of a MO resistor.
Another publication is surprising as well [Ver 1992]. Energy handling tests on commercially available MO resistors
are reported there. The resistors were of 22 mm height and 53 mm in diameter, they had a continuous operating
voltage of 2,5 kV and a nominal discharge current of 10 kA. Unfortunately, no absolute values of energy handling
capability is given, the information is limited to general tendencies which, however, are remarkable. While all
reports published so far indicate an increase of energy handling capability with current density, the contrary is the
case here: energy handling capability decreases with increasing current density. Compared with all other findings
and to date's knowledge, this has to be judged as an error, for which reason ever.
Darveniza et al. [Dar 1998] investigated the performance of distribution arresters under multiple impulse stress
which is motivated by the nature of lightning flashes5. They tested 21 porcelain housed arresters from six different
manufacturers. Additionally, they investigated MO resistors with different coating systems of one manufacturer, and
further MO resistors of other manufacturers in different surrounding media. Impulse currents 8/20 s at amplitudes
from 5 kA to 11 kA were applied, as well as 4/10 s impulses from 40 kA to 100 kA. The 8/20 s impulses were
multiple impulses at time intervals of (15150) ms. No information of the energy handling capability is published,
but an interesting observation is reported. The MO resistor stacks flashed over. Tests with surrounding gases
modified for different dielectric strengths (air at normal density, air at lowered pressures, SF6) resulted in the same
behavior. Thus, obviously, the flashover under this kind of multiple impulse is initiated not outside the resistor but
directly underneath or within the coating, a phenomenon that has also been observed for the new lightning current
impulse 90/200 s (see later in section 3.4.2).
Dengler [Den 1998] intensely investigated electrical degradation of MO resistors
(in terms of watt loss and leakage current increase) under an extremely high number (up to 400 impulses) of
lightning impulse current stresses at amplitudes around nominal discharge current and finally derived online
monitoring procedures from his findings. He investigated two different kinds of MO resistors (material A: height
46 mm, diameter 38 mm, nominal discharge current 5 kA; material B: height 40 mm, diameter 74 mm, nominal
discharge current 20 kA). Besides the impact of current amplitude and front steepness, he also investigated
recovery effects at different temperatures or time intervals between the individual impulses. As cause of electrical
degradation, he suggests migration of negative oxygen ions towards the inner region of the ZnO grains. This
changes the oxygen ion concentration at the grain boundaries and has effect on the barrier voltage. Recovery may
take place under certain conditions by negative oxygen ions travelling back to the boundaries.
Klein [Kle 2004] investigated changes of material properties by impulse currents, expressed by changes of leakage
current, reference voltage and power loss. He also looked very closely to fine cracks on the resistor surface. As a
good approach for generalization, he introduced a common reference current density of Jref = 0,12 mA/cm
(according to the standards, manufacturers are free to specify their reference current in any suited way).
Unfortunately, in case of asymmetries in polarity (which is typical after unipolar impulse current stress), he used the
average of positive and negative current amplitudes for the determination of the reference voltage, which makes
comparison of similar results difficult that are found according to the definitions in the standards.
The contributions cited so far dealt with the particular aspects of impulse energy handling capability. Only few
publications can be found on thermal stability issues. First studies were performed and published by [Lat 1983] [Lat
1985], where the thermal behavior and thermal stability limit of individual MO resistors and MO distribution
arresters was investigated and simulated by means of a transient network analysis. St.-Jean et al. [StJ 1990]
reported about a similar approach for high-voltage arresters up to 120 kV rated voltage, were the MO temperature
is approximately evenly distributed along the arrester axis as well. The problem becomes more complex, however,
if high-voltage arresters of several meters in height are considered, since they represent structures of distributed
parameters with a spatial distribution of all electrical and thermal quantities. In [Hin 1987], [Hin 1989] and [Hin
1990] approaches and performed simulations on such arresters are reported. They were also based on a
5
In 55% of all cases lightning flashes are composed of two or more individual strikes [And 1980].
Page 111
comparatively coarsely structured distributed parameter network. Though further progress has been made in this
field for instance [Hin 2008] reports about successful coupling of a three-dimensional non-linear electroquasistatic and a thermal field problem for UHV arresters till now really satisfying approaches to simulate thermal
stability limits under special consideration of the axial temperature distribution in HV-, EHV- and UHV arresters
have not been published.
In a second part of the program (in the work frame of Cigr WG A3.25) issues such as durability (impact of number
of impulses) and the problem of single versus multiple impulse stress will be addressed as well as statements on
failure risk (failure probability versus absorbed energy), and finally the impact of uneven axial temperature
distribution in high-voltage arresters on the thermal stability limits. This will allow deriving rated energies for specific
applications of surge arresters. Finally, better energy definitions and simpler test procedures for future revisions of
the arrester standards shall be derived and suggested.
Results were published so far in [Rei 2008a] [Rei 2008b] [Tuc 2009] [Hin 2009] and are including some new and
most recent results summarized below.
Page 112
It was for the first time, as well, that a systematical investigation with the new 90/200 s impulse ("lightning impulse
discharge" acc. to [IEC 2009]) was performed.
Page 113
Initial measurements
Uch,1 at Jch = 0,12 mA/cm (after 5 s)
Ures,1 at I = In
Impulse test
(energy injection)
Visual inspection:
mechanically failed?
yes
no
Measurement of characteristic voltage
Uch,2 at Jch = 0,12 mA/cm (after 5 s)
Exit measurements
95% U ch,1
105% Uch,1 ?
Uch,2
no
yes
Measurements at lightning current impulse
Ures,2 at I = In
Imd at J = 1,5 kA/cm
Visual inspection:
mechanically failed?
yes
no
no
95% Ures,1
105% Ures,1 ?
Ures,2
yes
OK
Figure 3.4: Flowchart of the test and evaluation procedure
Page 114
defect
Some of the energy test series for station class MO resistors were performed at alternating current stress. In this
case, due to restrictions of the test setup, voltage was applied until mechanical failure of the MO resistor.
Therefore, no initial or exit measurements had to be performed for this test series. After failure of the MO resistor
the short circuit current of the transformer was interrupted by a vacuum circuit breaker. Since the transformer short
current was flowing for a time of up to 40 ms, it was not in all cases possible to exactly determine the failure mode
of the MO resistor.
Page 115
solved in the near future. Actually, MO resistors have not been optimized for the extreme current stress of the
lightning current test as specified in Annex N of [IEC 2009].
Figure 3.5: Mean failure energy vs. amplitude of current density for Size 1 MO resistors
For make "T", tests at long-duration current impulse stress were also performed up to mechanical failure (rather
than to apply the "complex failure criterion"). It can be seen from Figure 3.6 that in this case the mean failure
energy follows the expected dependence of current density. This demonstrates that the differences in Figure 3.5
between the a.c. and the impulse current tests are not related to the different current shapes but only to the
different applied failure criteria. It further allows concluding that energy handling test can also be made with
alternating current stress if this is considered more convenient. The only difference to impulse current testing will
then be the lower failure energies due to the lower achievable current densities, which, however, can easily be
taken into account by correction factors.
Page 116
1800
Diameter 60 mm
Heigth 40..45 mm
1600
1400
1 ms
1200
2 ms
100 ms 4 ms
1000
8s
800
400
AC
200
0
0,1
10
100
1000
10000
Figure 3.6: Mean failure energy vs. amplitude of current density for Size 1 MO resistor of make T;
comparison of failure criteria until mechanical failure and complex for the long-duration current
impulse stress
The linear dependence between logarithm of current (density) and logarithm of time to failure could basically be
verified, see Figure 3.7 (for the impulse tests the prospective impulse time is used for the time scale; this is due to
the test conditions, as the test was performed with standard current impulses and not with an impulse current
lasting up to mechanical failure of the sample). It is interesting with this kind of depiction that for the short impulse
times and high current densities, respectively, the different failure mechanism of resistors "S" and "U" (flashover
instead of breaking) can clearly be identified by a change of the rate of rise of the curve but only if the curve is
carefully interpreted. One may also (erroneously!) conclude that the same linear log-log dependence is valid over
the full covered range, as the dramatically decrease of the energies at the left end of the curve looks quite
"harmless" in the logarithmic scale and can easily be ignored. In general, this way of depiction points out general
dependencies, but is too coarse for quantitative evaluations.
Page 117
Figure 3.7: Mean values of current density amplitude vs. time to failure
Figure 3.8: Mean failure energy vs. amplitude of current density for Size 2 MO resistors
Page 118
The results on the distribution MO resistors "Size 2" from six different manufacturers are shown in Figure 3.8. The
following can be derived from this picture:
1. For the long duration current impulses, mean failure energies are in the range of (6001000) J/cm, with
increasing values for increasing amplitudes of current density. These values can be directly compared with
those of Figure 3.5 which are typically in the range of (8001200) J/cm. The distribution MO resistors
thus have (1525) % lower energy handling capability. This is explained by the different dominating failure
mechanisms, as will be shown later in more detail. In general, the "Size 2" resistors would much more often
fail due to a change of the U-I-characteristic, which can only be found with the help of the "complex" failure
criterion.
2. Except for one make ("X"), beginning with a current density of several hundred A/cm, the mean failure
energy does not increase any more but even decreases, down to values of only (150650) J/cm at high
current impulse stress. This is, of course, again due to the application of the complex failure criterion. For
tests carried out up to mechanical failure an increase would have been expected in this range.
At this point some details about the typical failure mechanisms shall be given. Figure 3.9 shows the failure
mechanisms depending on the impulse shape for the station class MO resistors "Size 1", makes "S", "U" and "X".
Figure 3.10 gives the same information for the distribution MO resistors "Size 2", makes "S", "U", "V", "W" and "Y".
Figure 3.11 gives an idea about the meaning of failure mechanisms "cracking" (CR), external "flashover" (FO),
"puncture" (PU) and the special characteristic of a "flashover" in case of the 90/200 impulse current stress, which
originates from a puncture of the coating.
Page 119
100
100
80
80
60
60
40
40
20
0
CR
BR
FO
B
PU
DU
MF Uch
Uref Ures
Ures
20
4 ms
2 ms
1 ms
90/200 s
4 ms
2 ms
1 ms
90/200 s
0
CR
FO
PU
MF Uch
Ures
V
100
100
80
%
80
60
60
%
40
40
4 ms
20
4 ms
2 ms
1 ms
90/200 s
20
2 ms
1 ms
CR
BR
FO
B
DU
PU
CR FO
90/200 s
MF
Uch
Uref Ures
Ures
100
100
PU
MF Uch
Ures
80
80
60
60
40
40
4 ms
2 ms
1 ms
90/200 s
20
0
CR
BR
FO
B
20
DU
PU MF Uch
Uref Ures
0
CR
FO
PU MF
Uch Ures
4 ms
2 ms
1 ms
85/180 s
Failure mechanisms:
CR Cracking
FO Flashover
PU Puncture
Page 120
100
100
80
80
60
60
40
40
20
4 ms
1 ms
4/10 s
Uch Ures
Uref
Ures
FO
1 ms
PU
MF
Uch
U ref
MF
4/10 s
Ures
PU
DU
MF
FO
B
BR
CR
CR
BR
4 ms
0
DU
20
Ures
100
100
80
80
60
%
60
40
%
40
20
0
CR
BR
B
FO
DU
PU MF Uref
Uch Ures
Ures
4 ms
1 ms
4/10 s
20
0
CR
BR
B
FO
DU
PU
MF Uref
Uch Ures
4 ms
1 ms
4/10 s
100
100
80
80
60
60
40
40
4 ms
2 ms
1 ms
90/200 s
4/10 s
20
0
CR
FO
PU
MF Uch Ures
20
2 ms
0
BR
CR
4/10 s
B
FO
DU
PU
MF
Uref
Uch Ures
Ures
Failure mechanisms:
CR Cracking
FO Flashover
PU Puncture
Page 121
Figure 3.11: Failure mechanisms (from left) cracking (CR), flashover (FO), puncture (PU) and
flashover (FO) in the special case of 90/200 impulse current stress
For the "Size 1" MO resistors, main failure mechanism of make "S" is change of the characteristic voltage and
cracking; only at the 90/200 s impulse, it mainly failed by flashovers. For make "X" the dominating failure
mechanism is cracking, and even in case of the 90/200 s impulse it is much more the change of the characteristic
voltage than flashover. It is also interesting that in general puncture obviously is not a common failure mechanism,
a finding that is in contradiction to what was published in [Rin 1997] (see section 3.4.1). Another important
observation is that change of the residual voltage (by more than 5 %) is not a concern at all. This criterion is often
used in the arrester standards and has to be questioned for further revisions.
For "Size 2" MO resistors, the dominating failure mechanisms are change of the characteristic voltage and
flashover. This is not only the case for 4/10 s and 90/200 s, where this might have been expected, but also for
the long duration current impulse stress in one example. Particularly for the "Size 2" resistors one can identify a
typical failure mechanism pattern of a certain make. Here again, change of the residual voltage does not take
place, and puncture is not a relevant failure mechanism, either.
Figure 3.12 and Figure 3.13 give detailed information about the change of the characteristic voltage for the "Size 2"
distribution voltage MO resistors, depending on the 4/10 s impulse current peak value and the related injected
energy.
Current peak value in A
5
40.000
0
60.000
80.000
100.000
120.000
140.000
160.000
180.000
200.000
220.000
Change of U ch in %
-5
-10
S
U
V
W
X
Y
-15
-20
-25
-30
-35
-40
Figure 3.12: Change of characteristic voltage vs. 4/10 s impulse current peak value
Page 122
Energy in J/cm
5
0
100
200
300
400
500
600
700
800
900
1.000
Change of U ch in %
-5
-10
S
U
V
W
x
Y
-15
-20
-25
-30
"Size 2"
Diameter 40 mm
Height (30..40)
-35
-40
Figure 3.13: Change of characteristic voltage vs. energy injection by 4/10 s impulse current
The differences among the different makes of MO resistors are impressing. Only two of them "V" and "W"
exhibit a change of less than 5 % in the characteristic voltage after one 100 kA current impulse 4/10 s, which is
the standard test impulse during the operating duty test on distribution arresters according to [IEC 2009]. The other
three makes are at 5 % and at 15 % for this current amplitude. Interesting as well is the fact that some makes
are obviously optimized for minimum impulse current degradation. "V" reaches the 5 % limit at a current of 150
kA, and "W" has a decrease in the characteristic voltage of only 7 % at a current amplitude of 190 kA, the slope of
the curve being extremely flat. From Figure 3.12 it can be seen that all investigated resistors easily reach energy
handling values (without mechanical failure) of (700900) J/cm assumedly even higher, but the impulse current
generator was at its limits. This example demonstrates how important the introduction of a "complex" failure
criterion is. It is acknowledged that all of the investigated MO resistors will perform well in a complete distribution
arrester because the decrease of the characteristic voltage (for practical applications this might be the reference
voltage) of (5.15) % at 100 kA current amplitude is taken into consideration for the dimensioning of the arresters
all of them are designed to pass the operating duty test. However, deterioration of the material has definitely
taken place (showing a linear dependence from the current amplitude), and it may just be discussed if, for instance,
a change of 10 % in the characteristic voltage should be used in the complex failure criterion rather than 5 %.
This would result, e.g. for make "Y", in an increase of the failure energy from actually 200 J/cm to 300 J/cm, see
Figure 3.13.
For the first time, mechanical shock waves in MO resistors under high current impulse stress, so far only
theoretically predicted by simulations (see section 2.2), could be measured in this investigation. Figure 3.14 shows
the calculated temperature increase acc. to Equ. 3.1 in an MO resistor of 40 mm diameter and 45 mm height under
100 kA high current impulse stress and under assumption of a homogeneous temperature distribution in the
material. Temperature increases by nearly 110 K within a time of 12 s. This adiabatic step increase causes
extreme thermo-mechanical stress, and mechanical shock waves will travel through the material.
Page 123
140
350
Temperaturanstieg
in K
Temperature increase
Current
in kA
Strom
Energy in J/cm
Energie
120
300
100
250
80
200
60
150
40
100
20
50
Energie in
in J/cm
J/cm
Energy
Temperaturanstieg
in kAin kA
Temperature
increaseininKK/ Strom
/ Current
0
-2
10
12
14
Zeit in s
s
Time
7000
140
6000
120
5000
100
4000
80
3000
60
Kraft
Forcein N
Strom
Current
in kA
2000
Strom in
Current
inkA
kA
Force
kN
Kraft in N
Figure 3.14: Calculated energy injection and temperature increase under 100 kA high current impulse
stress on an MO resistor of 40 mm diameter and 45 mm height
40
1000
20
0
-10
10
20
30
40
50
60
70
80
90
Zeit in
Time
ins
s
Figure 3.15: Current and force measured on an MO resistor 40 mm diameter and 45 mm height
The result of an actual force measurement on such MO resistor under this kind of stress is shown in Figure 3.15. It
must be noted, however, that the diagram shows the uncompensated output signal of the force sensor, which may
be affected by the test setup, especially the supporting structure and the force sensor itself. Therefore, the result
has to be carefully interpreted and the general validity to be further verified.
The propagation speed c of an acoustic shock wave in MO ceramics can be calculated as
100 GPa
5420 kg/m
4300
m
s
4, 3
mm
s
(equation. 3.2)
where E is the module of elasticity and is the density (values taken from [Len 2000]. Thus for MO resistors of 27,8
mm height and of 37 mm height, respectively, especially cut to these heights for this investigation, the required time
for travelling two times along the height would be 12,9 s and 17,2 s, respectively. This is quite well correlated
with the comparative measurements shown in Figure 3.16. These investigations will be continued.
Page 124
120
Strom in
Current
in kA
kA
100
80
60
40
20
0
-20
-5
10
15
20
25
30
35
40
35
40
9
17,3 s
Force
in kN
kN
Kraft in
9,1 s
Hhe 37
Height
37 mm
mm
Height
27,8 mm
mm
Hhe 27,8
11,6 s
6
5
4
6,3 s
3
-5
10
15
20
25
30
Time
s
Zeit inins
Page 125
99,8
99,5
99
98
95
95
90
90
Wahrscheinlichkeit in %
99,8
99,5
99
98
80
70
60
50
40
30
20
Probability in %
Probability in %
Wahrscheinlichkeit in %
Mepunkte
Measuring
points
Treppenkurve
Step curve
Verteilungsfunktion
function
Distribution
Vertrauensbereiche
Verteilung
Conf.
interv. distribution
Vertrauensbereiche
Quantile
Conf.
interv. quantile
80
70
60
50
40
30
20
10
10
5
Verteilungsfunktion
function
Distribution
2
1
0,5
Verteilung
Vertrauensbereiche
Conf. interv. distribution
Vertrauensbereiche
Quantile
Conf. interv. quantile
0,2
1470
1570
1670
1770
2
1
0,5
0,2
1870
810
860
910
Energie in
in J/cm
Energy
J/cm
960
1010
1060
Energiein
in J/cm
Energy
J/cm
Figure 3.17: Examples of statistical evaluations (Normal Distribution) of impulse current tests (left) and
alternating current tests (right) with 95 % confidence intervals
1,8
1,6
1,4
1,2
1
0,8
0,6
0,4
0,2
6 cm
8 cm
10 cm
0
0
10
12
Figure 3.18: Mean failure energy (incl. standard deviation) vs. current density amplitude for MO
resistors of same make and same height but different diameters
Since tests at alternating current up to mechanical failure can be made at comparatively low effort, this procedure
might be given preference in certain cases. Figure 3.18 shows an example. Here, the objective was to compare the
influence of the MO resistor diameter on energy handling capability. Each measuring point represents energy tests
at the given current density amplitude on about 50 MO resistors of the given diameter. Such test program can be
performed in comparatively short time.
Page 126
The interesting finding here is not that failure energy increases with current density (which has been expected) but
that it decreases with diameter. This can only partly be explained by statistical (volume) effects as it is
demonstrated in the following example:
The probability P (W') of failure at a given specific energy W' for n MO resistors in parallel, each having a
probability p(W') of failure at the same specific energy is
P W'
1 p W'
(equation 3.3)
p W'
1 P W'
(equation 3.4)
where p(W') is the required failure probability of an individual MO resistor at the specific energy W' in order to
achieve the overall failure probability P(W') when a number of n resistors are connected in parallel. The number n
of parallel resistors can also be a volume factor of a larger diameter resistor.
When looking at the mean (or 50%) specific failure energy (where p = 0,5) of a resistor of 60 mm diameter, the
same specific energy injected into a resistor of the same make, but of 100 mm diameter (i.e. 2,8 times larger
volume at the same height; or n = 2,8), the failure probability acc. to equation 3.3 would be P = 0,856. If, vice versa,
the failure probability of the larger resistor shall be P = 0,5, the related failure probability of a 60 mm resistor
(volume factor = 1/2,8) would be, acc. to equation 3.4, p = 0,22.
From Fig. 3.16 (right) the 22% specific failure energy is approximately 4% lower than the 50% specific failure
energy. One might thus expect that a 100 mm diameter resistor has an approximately 4% lower mean specific
failure energy than a 60 mm resistor of the same make. Fig. 3.17, however, shows that the difference is much
bigger, i.e. in the range of 10%.
Therefore, the decrease of specific failure energy with diameter cannot be explained by statistical (volume) effects
alone. An additional influence may be the (in)homogeneity of the material. The larger the diameter the more difficult
it is to achieve homogeneity. However, the effect is not too much pronounced and particularly covered by the wide,
overlapping deviations from the average values. It may, anyway, be concluded that a diameter of 60 mm evidently
represents a kind of optimum where change of the U-I-characteristic under high current densities (compare Figures
3.5 and 3.8) and the effects of material in homogeneities at low current densities both have minimum impact on
energy handling capability.
Another outcome of this investigation is shown in Figure 3.19. Typically, in a test series of several hundred
specimens with energy injection up to mechanical failure there will be one or more "outliers", i.e. MO resistors that
fail at extremely low energy levels. Such performance of a batch was found for all investigated makes of MO
resistors, and it shows that there will always remain a certain unavoidable risk when going to the limits of specified
energy handling capability. This is less a concern for standard applications, where an arrester is made up from
comparatively few MO resistors and is very likely never stressed to its limits, but it is definitely an issue for the large
arrester banks for overvoltage protection of series capacitors. These outliers, by the way, can better be found with
this way of testing, i.e. test at energy stress up to mechanical failure. In an impulse test with a standard longduration current impulse, where the outcome of each individual energy stress would be only be "passed" or "failed",
the information about the low failure energy of sample number 32 in Figure 3.19 is usually not available (unless the
actual failure energy during each impulse is measured and evaluated).
Page 127
1200
Energie
in J/cm
Failure
energy
in J/cm
1000
800
600
400
200
0
1
13
17
21
25
29
33
37
Test
and sample number
Versuchsnummer
41
45
49
53
Figure 3.19: Example of failure energies at alternating current tests up to mechanical failures
3.4.2.5 Conclusions and Outlook
In this experimental investigation a statistically significant number of station and distribution class MO varistors from
eight well established American, European and Japanese manufacturers was tested to probe the energy limits for
different impulse and alternating current stresses. A major part of the test program is concluded and some general
observations and conclusions can right now be formulated:
-
Compared to an earlier comparative study [Rin 1997] an increase of up to 70 % and in average of at least
20 % in energy handling capability can be seen for the materials studied here (only the alternating current
tests are directly comparable, where the same failure criteria were applied), despite the fact that MO
resistors of larger (about two times) volume were tested, which in doubt would result in lower energy values
due to volume effects. This probably reflects the continuous improvements industry has made over the last
decade in processing, material formulation and block design. And evidently, cost pressure on the market
has not resulted in lower qualities.
So far no new, emerging suppliers of MO varistors could be included in the study. However, recent
references (e.g. [He 2007]) indicate that substantial differences might exist in the performance of such
materials, which should be studied further.
For the first time a "complex" failure criterion was introduced, which allows very sensitive evaluation of
different failure mechanisms. Since also changes of the U-I-characteristic are evaluated it allows
considering the impact of impulse energy stress on thermal stability issues. Some aspects of this failure
criterion might be further discussed, e.g. which alterations of the U-I-characteristic may be accepted as a
pass criterion, but basically this procedure has proven to be effective. Its application is, however, time
consuming.
For the first time such a comparative study included MO resistors designed for distribution and station
arresters. They differ mostly in their geometrical dimensions (cross-section, aspect ratio) and also show
distinctive differences in their performance, presumably due to the fact that different features are optimized
for the two different applications.
Energy stresses caused by fast impulses 90/200 s as they are specified for line arrester applications in
[IEC 2009] were as well evaluated for the first time in this study. Flashover or significant alterations in the
Page 128
U-I-characteristics show up as the major failure modes when reaching the energy limits under these fast
impulses. It has thus been found that the rule "increase of failure energy with higher current density"
(published by [Rin 1997] for the first time and basically confirmed by this investigation) must be confined,
as at a certain current density stress level the performance of the coating system may become dominant,
or in other words: it has not been optimized for this kind of stress. Practical implications of these findings
for specific applications in medium or high voltage arresters have to await further discussion.
-
Tests at alternating current stress have turned out to be suited for fast comparison of material properties
(benchmark tests), when the "complex" failure criterion shall not be applied. They allow statements on
failure probabilities down to approximately 1 %, which is not possible by impulse current tests. But it is still
an open problem how to reliably specify energies that would result in 0,1% failure probability or less.
MO resistors of approximately 60 mm diameter may constitute an optimum with regard to impulse energy
handling capability.
The change in residual voltage after energy absorption is not an issue and should be removed as failure
criterion in the standards. Change of power loss would be a sensitive failure criterion but is not sufficiently
reproducible. Change of the reference voltage could be considered for this purpose, instead.
One final general remark is important: all comparisons among different makes presented in this study are based on
mean failure energies (50 % failure probability). For these figures the statistical confidence is high, and they can
thus preferably be used for benchmark purposes. The mean failure energies, however, should not be mixed up with
the design energies, which are in the range of only 200 J/cm and thus far below the values reported here. No
reliable and serious statement can be given on, e.g., a 0,1 % failure energy, for reasons explained before, and
there is definitely no basis for any comparison of the investigated makes at such low probability failure energy
values. One should therefore be very careful with rash conclusions that a certain make might be "better" or "worse"
than another one, only based on this study. All investigated samples came from well-established manufacturers of
excellent reputation, and it is well known that failure rates of real arresters in the systems are only a few percent in
distribution and close to zero in transmission applications.
As the next steps the ongoing measurements will be completed. Thereafter, further studies might be necessary on
topics such as multiple impulse stress, durability, and impact of aspect ratio, probabilistic aspects in energy
handling, high gradient MO resistors or low performance makes.
Page 129
tolerances of the impulse current parameters. An improvement therefore would be to also state charge transfer (in
As) or specific energy in kJ per kV of rated voltage.
The high current impulses during the high current impulse operating duty test on distribution arresters and lightduty station arresters (Cl. 8.5.4 of [IEC 2009]) represent a considerable energy stress for this class of arresters, but
are not a good basis for an energy handling capability specification either as the injected energy may vary strongly
when utilizing all allowed tolerances. Historically, this test has been introduced primarily as a dielectric withstand
test for the gapped SiC arresters, rather than an energy handling test.
This situation in the standards may have been sufficient for "classical" arrester applications. But with state-of-theart MO arresters many new applications have become possible and usual. Examples are line arresters, arrester
banks for protection of FACTS (particularly series compensation capacitors), shunt capacitor and reactor
protection, arresters in HVDC applications and others. Their application require a more sophisticated consideration
of energy handling capability, and an increasing number of users has profound knowledge about the arising energy
stresses and specify requirements on energy handling capability. But today's definitions and tests do not give
sufficient guidance in this respect. Therefore, activities have been started in IEC as well as in IEEE to improve the
arrester standards in this respect.
LD class
Virtual
duration of
peak T, in s
Charging voltage
UL, in kV (d.c.)
4.9 Ur
2000
3.2 Ur
2.4 Ur
2000
3.2 Ur
1.3 Ur
2400
2.8 Ur
0.8 Ur
2800
2.6 Ur
0.5 Ur
3200
2.4 Ur
Page 130
LD class
U s (kV)
245
300
420
550
800
Table 3.2:
Recommended line
discharge classes
depending on system
voltage
Figure 3.20: Specific energy injection (by two consecutive line discharges) during switching impulse
operating duty test in kJ/kV of rated voltage dependent on the ratio of switching impulse residual
voltage U res to the r.m.s. value of the rated voltage U r of the arrester (in accordance with [IEC 2009])
As no direct conclusions about the energy stress can be drawn from this table, Figure 3.20 depicts the converted
energy in a test object during the switching impulse operating duty test (injected by two line discharges), with
reference to its rated voltage. This energy is not a fixed value, but depends on the arrester's switching impulse
residual voltage. The higher the residual voltage, the less energy the arrester absorbs during the line discharge.
With the help of Figure 3.20 a typical problem related to the LD class definition shall be explained. If a design is
applied with a given amount of specific thermal energy handling capability, then the arrester can, depending on its
actual residual voltage, be assigned to different LD classes, as shown in the following example (red dashed lines in
Figure 3.20): when using a design that can absorb an energy of 4 kJ/kV during the operating duty test, the arrester
is of LD 2 at a ratio of Ures/Ur = 2. However, it can also be assigned to LD 3 at the ratio of Ures/Ur = 2,35. But the
apparently "better" LD 3 arrester might possibly be worse for the intended application, since its protective level is
higher. In order to reach LD 3 while maintaining a ratio of Ures/Ur = 2, a design must be used with a thermal energy
handling capability of almost 6 kJ/kV, as indicated by the blue dotted lines in Figure 3.20, that would mean
application of MO resistors with greater diameters. Inversely, one can only draw conclusions from the LD class in
connection with the residual voltage as to the (thermal) energy handling capability of an arrester, and thus about
the used MO resistors.
For standard applications, one can simply count on recommendations in the application guide [IEC 2000], based on
the system voltage level (Table 3.2). In practice, however, users often tend to select the next higher LD class,
respectively, in the table. That leads to the problem that the current highest LD 5 can frequently not meet the
demands of the extra-high-voltage systems with Us > 550 kV. In fact, at this voltage level, and sometimes even at
the 550-kV-level itself, MO resistor diameters and/or parallel connections of resistors are used, which have much
higher energy handling capability than specified by LD 5. This is also a particular problem of the emerging 800 kV
d.c. and 1100/1200 kV a.c. applications, where specific energy handling values of (2550) kJ/kV of rated voltage
will be required at switching impulse protective levels in the range of only 1,85 or even less [Ric 2007]. It can easily
be seen from Figure 3.20 that LD 5 is by far not sufficient for these applications: an LD 5 operating duty test will
inject only (2550) % of the required energy. For UHV arresters, the test procedure has therefore to be modified.
For instance, energy could be injected by more than two long duration current impulses. It could also be discussed
if other test parameters, such as the time duration of the impulses, may be changed. Summarizing, at least three
work items can be identified for a future revision of the standard: the possible replacement of the LD system by a
purely energy based rating system, clear definitions and differentiation of different kinds of energy handling
capability, and specification of higher energy handling values than today and the related test procedures.
Page 131
The relevance of a number of eighteen current impulse applications is questionable: eighteen impulses
each on three samples results in 54 energy stresses. This, even if identical characteristics of each of the
three samples are assumed in the best case, cannot validate a failure probability of less than
p = 1/(318) = 0,0185 or 1,85 %, respectively, a value totally insufficient for real arrester applications, see
also section 3.3.2. Furthermore, it does not seem sufficient to specify such test as type test only, as energy
handling capability is also or distinctly a matter of production quality.
As the test sample temperature has effect on the test result, the thermal conditions of the test setup and
the test procedure should be better specified.
The injected energy per line discharge is usually less than the long-duration current withstand values
specified by the manufacturers; higher values should possibly be specified. It must be noted, however, that
not all manufacturers clearly specify how the long duration current withstand values are defined and
determined; therefore, comparisons with the line discharge currents have to be made with reservations.
The maximum allowed change of 5 % in residual voltage after the test should be subject to discussion, as
such extreme changes are only seldom observed, while other relevant regions of the voltage-currentcharacteristics (e.g. reference voltage) react more sensitively to impulse degradation; see also section
3.2.5.
While in case of 10-kA- and 20-kA arresters the long-duration current impulse shall be a line discharge as
defined by the parameters shown in Table 3.1, for 2,5-kA- and 5-kA arresters (i.e. mainly distribution
arresters), the relevant current impulses shall be only (50 A / 500 s) and (75 A / 1000 s), respectively.
These values are in fact too low to inject any notable energy even into a light-duty distribution arrester. The
applied MO resistors for these arresters are usually specified for much higher values to date, e.g. in the
range of (200 A / 2 ms). Therefore, this requirement should be discussed as well.
Page 132
Procedures a) and b) are covered quite well in the actual version of the standard, latest since the accelerated aging
test has been extended to the case that part of the arrester is stressed by voltages higher than the reference
voltage of the MO resistors. Item c), however, is questionable: the application of twenty lightning current impulses,
superimposed to an applied power-frequency voltage, originates from the era of gapped SiC arresters and served
for preconditioning of the series gaps, which had to interrupt the power-frequency follow current each time. For
today's MO arresters the test procedure in its actual version seems meaningless, and as this test is particularly
difficult to perform it should be simplified in the future by removing the requirement for a superimposed powerfrequency voltage.
The operating duty tests shall be performed on prorated arrester sections that represent the electrical and thermal
characteristics of the full arrester (Cl. 8.5.3.2 of [IEC 2009]). But the requirements on these sections are partly
contradictory, in that thermal equivalence is required on one hand and use of the same material and dimensions for
the housing as for the real arrester on the other. A revision of this part of the standard has therefore to make the
requirements more consistent. Furthermore, since the high-current impulse applications also impose mechanical
and dielectric stress to the sample, presence of the mechanically supporting structure should be required.
Page 133
procedure could not be simplified by allowing to inject the energy by any kind of long duration current impulses
without requirements on virtual duration of the peak and amplitude, as long as the energy requirements are fulfilled.
For instance, the test could in general be performed with two or even more (in case of extra-high energy injection
requirements) rectangular impulses of variable amplitudes. The main uncertainty to answer this question namely
the different reactions of the test sample expected for different virtual durations of the peak and different current
densities, respectively has been clarified by the energy handling research program reported in section 3.4.2.
However, if more than two impulses shall be allowed for energy injection, possible cooling of the sample between
the individual impulses should be taken into account by correction factors for the required energy input.
3.5.2.4 High current impulse operating duty test on high lightning duty arresters
This test according to Annex C of [IEC 2009] is intended to be performed only on 20-kA high lightning duty
arresters, especially applicable for high lightning density areas with highest system voltages in the range
1 kV Us 52 KV. The energy stress is imposed by three impulse currents 30/80 s of 40 kA peak value, one
minute apart without cooling. After the third impulse thermal stability has to be verified as in the other operating
duty tests. No published information is available about the severity of this test compared with the switching surge
operating duty test. Obviously, there is only little need for this test as this kind of arrester seems to be quite
uncommon. And in fact, Annex C is unknown to many of the users. It may thus be asked if it can be totally removed
in a future revision of the standard.
for any of the three test samples. Energy and charge values are taken from tables with standard values. These
tables give steps of rated energy values up to 20 kJ/kV(Ur) and charge values up to 10 J (for comparison: a typical
LD 5 arrester has a thermal energy handling capability of about 13 kJ/kV(Ur) and a single impulse (2 ms) energy
handling capability of about 8 kJ/kV(Ur)). For transmission line arresters, which are commonly of LD classes 2 to 4,
Figure 3.21 shows the current amplitudes which inject the same energy to the sample as the respective line
discharge, resulting in peak values up to about 15 kA (and two times this value if the same energy shall be injected
as by two line discharges in the operating duty test).
Page 134
16
14
12
10
8
6
4
2
0
LD 2
590 A / 2000 s
LD 3
720 A / 2400 s
LD 4
880 A / 2800 s
Figure 3.21: Required 90/200 impulse current amplitudes for injecting the same energy into an MO
resistor as by one line discharge equivalent to LD classes 2 to 4
It is a new approach of the IEC arrester standard to specify an arrester's energy handling capability by three
different parameters. Background is that, depending on application and system voltage level, charge transfer
capability, energy handling capability or current carrying capability may be of higher importance. As gapless line
arresters represent a comparatively new application it is actually difficult to make a commitment exclusively for one
of these parameters. Thus this set of three parameters has found international consensus. Considering the fact that
energy handling capability of MO resistors depends on current density and impulse time duration, respectively, this
new test procedure is a reasonable approach to verify TLA performance under short-duration, high-amplitude
current stress. The results of the energy handling research program clearly indicate that there are distinct
deviations from the expected energy handling capability for this current impulse (see, e.g., Figure 3.5). Together
with the existing long-duration current test two important extremes of current parameters are thus covered by the
standard.
Page 135
requirements. Instead, a fixed current amplitude and a fixed time duration are chosen, which in most cases lead to
higher stress than the standard long-duration current impulse withstand test. It has just been addressed and
motivated in section 3.5.1 that a charge transfer capability or a specific energy handling capability would be a better
definition for this purpose.
Anyway, it cannot be excluded that other manufacturers specify, for instance, a single impulse energy handling
capability with the same term (probably leading to higher current values). Thus the problem with this kind of nonstandardized energy handling definition becomes evident: the test procedure is not specified and in many cases
not even explained in the manufacturers' catalogues, the time duration of the current impulse may be chosen to
different values, and any current value may be specified, leading to a "battle of catalogue values". The actual
situation is thus not satisfying, and such kind of a well-defined energy impulse handling specification should
officially be adopted in the standard and become mandatory.
IEC 60099-4
2,5 kA
5 kA
10 kA, LD 1
Intermediate
5 kA, LD 1 or 2
Station 10 kA
10 kA, LD 3 or 4
Station 15 kA
20 kA, LD 4 or 5
Station 20 kA
20 kA, LD 4 or 5
Table no. 3.3: Comparison of arrester classes acc. to IEEE and IEC standards
3.5.3.1 High current short duration withstand test
This test (clause 8.12 of [IEE 2005]) has to be performed on a complete arrester or an arrester section, i.e. the
housing is included. Two high current impulses shall be injected, followed by application of power-frequency
recovery voltage6 for at least thirty minutes. The intention is to demonstrate the design's dielectric strength as well
as thermal stability. This may be well achieved for distribution arresters, which are tested at current amplitudes of
40 kA, 65 kA or even 100 kA. But it imposes neither a notable dielectric nor energetic stress on intermediate and
station arresters that have to be tested at amplitudes of 65 kA only. For these arresters, the test seems
meaningless. Also the following is doubtful:
the current wave shape shall be 4/10 s (0/+50%), which causes the same problem of an undefined
energy injection level as in the high current impulse operating duty test of [IEC 2009];
no tolerance at all is given for the current amplitude, which is unrealistic for practical testing;
Voltage that causes the same watt losses in the actual MO resistors as a voltage equal to MCOV would do in aged resistors of
the same make and of the highest specified watt losses.
Page 136
the allowed time interval of five minutes between impulse application and energization at power-frequency
voltage is too far away from real service conditions; thermal stability cannot really be demonstrated by this
test procedure.
the test generator must be very carefully adjusted for each transmission line discharge level, making
testing rather complicated (the IEC standard, instead, only requires that the energy and time values
resulting from the test parameters given in Table 3.1 are met, which allows less precise adjustment of the
test generator);
placing the impulses in three groups of six applications each may lead to unnecessarily severe (and
unrealistic) thermal stress to the MO resistors;
the requirements for system voltage levels above 400 kV are to weak, leading to the strange situation that
an arrester in a 362 kV system is specifically less stressed than an 800 kV arrester; reason is that for EHV
systems the charging voltage of the line has been set to only 2,0 p.u. of the system voltage, whereas for
the lower system voltage levels 2,6 p.u. are assumed;
a time interval of five minutes between injection of the last impulse and power-frequency application is too
long
Page 137
Page 138
4. Summ ary
Author in charge: Bernhard Richter
MO arresters are nowadays installed in all kind of electrical power systems from low voltage up to UHV. They are
intended to protect equipment and installations against overvoltages. Due to the various applications the MO
arresters have to withstand severe stresses from the electrical system, from lightning and from ambient. In this new
TB the different types of stresses are listed and severe stresses, e.g. winter lightning, seismic stresses and severe
pollution of polymeric housings, are shown and examples for test procedures are given.
Main focus is given on the progress in arrester technology and application in the past 20 years.
For understanding the interaction of the modern MO surge arresters with the system conditions and the ambient
stresses, the basics of the MO material and the various designs of MO arresters on the market are given in detail.
The working group engaged in a critical review of the applicable standards and initiated a research program with
international participation on energy handling withstand capability of MO resistors and arresters. The results show
that the impulse withstand capability has increased by app. 20% for the manufacturers participating in this study,
compared to previous investigations. However, differences exist in the mean failure energy and the failure modes
depending on the type of impulse stress. Unfortunately, materials from emerging countries were not available for
this study when it was initiated.
For clarity it should be noted, that the observed mean failure energy of the MO resistors is three to four times
higher than the design energy of the MO arresters as proved in the relevant operating duty tests. This gives a good
safety margin and confidence in the todays designs and materials. However, with increasing system voltage the
number of MO resistors easily reaches several hundreds of blocks per arrester, and then statistical evaluations
may have to be considered.
For a more sensitive evaluation of the MO resistors in this study, a new complex failure criterion was developed
and used, including the change of the electrical characteristic of the MO resistors in addition to the simple failure by
cracking or flashover.
The different types of failure modes depending on the different current wave shapes can lead to further
improvements of the MO resistors.
The results from the research project and the review of the existing standards, together with new applications,
leaded to a new classification of the energy capability of MO surge arresters. The new classification concept,
charge transfer classes instead of line discharge classes, is introduced in Rev. 3.0 of IEC 60099-4. The same
concept is adapted for instance in EN 50 526-1, which is a test standard for MO surge arresters to be used in d.c.
traction systems.
MO arresters are applied more and more for insulation coordination reasons and not only for protection of a single
high voltage equipment against overvoltages. This is especially the case for UHV a.c. and d.c. systems. The
development of MO resistors and arresters is ongoing with the goal of size and cost reduction but in the same time
keeping the high quality and reliability. This leads to the development and use of MO resistors with increased field
strength to reduce the size of the complete design, e.g. in GIS applications. Further on, with the very tall arresters
for UHV systems, which are easily taller than 10 m, the question of how to test complete arresters comes up.
Simulations may help to reduce testing. These questions and ongoing research on the energy withstand capability
of MO resistors and arresters are dealt with in working group A3.25 of SC A3 and will be published in a separate
Technical Brochure.
Page 139
APPENDIX 1
Following Technical Brochures of Cigr are dealing with surge arresters and their application:
TB 60 Metal Oxide Arresters in AC Systems
by WG 06 of SC 33, 1991
TB 287 Protection of MV and LV Networks against Lightning. Part 1: Common Topics
by CIGRE-CIRED JWG C4.4.02, 2006
TB 441 Protection of MV and LV Networks against Lightning. Part 2: Lightning protection of Medium
Voltage Networks
by CIGRE-CIRED JWG C4.4.02, 2010
TB (XXX) Protection of MV and LV Networks against Lightning. Part 3: Lightning protection of Low-Voltage
Networks
by CIGRE WG C4.408, to be published 2013
TB 440 Use of Surge Arresters for Lightning Protection of Transmission Lines
by CIGRE WG C4.301, 2012
TB 455 Aspects for the Application of Composite Insulators to High Voltage ( 72 kV) Apparatus
by CIGRE WG A3.21, 2011
This TB addresses the special case of surge arresters with composite insulators.
Page 140
References
[AIE 1950]
[And 2000]
H.Andoh, S.Nishiwaki, H.Suzuki, S.Boggs. J.Kuang, IEEE Electrical Insulation Magazine, Vol.16(1),
pp25 (2000) and H.Andoh, Y.Itoh, S.Nishiwaki, T.Imai, S.A.Boggs, J.Kuang, J.Soc.Mat.Sci.(Japan),
Vol.49(9), pp982 (2000)
[And 2003]
[AS 1996]
AS 1307.2-1996: Surge arresters Part 2: Metal-oxide surge arresters without gaps for a.c. systems
(Australian Standard)
[Asa 1994]
A. Asakawa et al. The discharge progress aspect and the performance of current shape of winter
lightning CRIEPE Report, T93024 (in Japanese), 1994
[Bal 2004]
[Bar 1996]
[Bar 1996]
M.Bartkowiak, M.G.Comber, G.D.Mahan, J.Appl.Phys., Vol79(11), pp8629 (1996) and IEEE Trans.
Power Delivery, Vol.14(1), pp152 (1999) and ibid. Vol.16(4), pp 591(2001)
[Bar 1996a]
Bartkowiak, Comber, Mahan, Energy handling capability of ZnO varistors, Journal of applied
Physics, Volume 79, Page 8629-8633
[Bar 1996b]
Bartkowiak, Mahan, Modine, Alim, Influence of ohmic grain boundaries in ZnO varistors, Journal
of applied Physics, Volume 79, Number 1, Page 273-281, 1996
[Bar 1996c]
Bartkowiak, Mahan, Modine, Alim, Lauf, McMillan, Voronoi network model of ZnO varistors with
different types of grain boundaries, Journal of applied Physics, Volume 80, Number 11, Page 65166522, 1996
[Bar 1999]
Bartkowiak, Comber, Mahan, Failure modes and energy absorption capability of ZnO varistors,
IEEE Transactions on power delivery, Volume 14, Number 1, Page 152-162, 1999
[Ber 1975]
K. Berger, R. B. Anderson, H. Kroniger, Parameter of Lightning Flashes, ELECTRA, No.41, pp. 2337, 1975
[Ber 2008]
S.Bernik, M.Podlogar, N.Daneu, A.Recnik; Materials and technology, 42(2), pp69 (2008)
[Bin 1993]
[Bir 1997]
D.S Birelli et al. EPRI Energy testing of transmission line surge arresters Cigr SC 33 Int.
Colloquium Toronto Canada 1997.
[Bla 1986]
G.Blatter, F.Greuter, Phys.Rev.B, Vol.33(6), pp3952 (1986) and Vol.34(12), pp8555 (1986) and
G.Blatter, D.Baeriswyl, ibid., Vol.36, pp6446 (1987)
[Bog 2000]
Boggs, Kuang, Andoh, Nishiwaki, Increased energy absorption in ZnO arrester elements through
control of electrode edge margin, IEEE Transactions on power delivery, Volume 15, Number 2,
Page 562-568, 2000
[Bog 2000]
S.Boggs, J.Kuang, H.Andoh, S.Nishiwaki, IEEE Trans. Power Delivery, Vol.15(1), pp128 and
Vol.15(2), pp562 (2000) and Conf. Rec. 1998 IEEE Int.Symp.Electr.Insul., pp464 (1998)
[Bue 2008]
[Cab 2004]
Page 141
[CAN 2004]
CAN/CSA-C233.1-87 (reaffirmed 2004): Gapless Metal Oxide Surge Arresters for Alternating
Current Systems, Electric Power Systems and Equipment, Canadian Standard
[Car 2003]
[Che 1996]
[Che 2002]
Q.Chen, J.He, K.Tan, S.Chen, M.Yan, J.Tang, Sci. in China (Series E), Vol.45(4), pp337 (2002); Y.Tu
et.al., IEEE proc. 8th Intern. Conf. Prop. and Applic. Dielectr. Mater., pp95, pp335 and pp959
(2006); J.He et.al. IEEE Trans. Power Del., Vol.19(1), pp138 (2004)
[Chi 1998]
Y.M. Chiang, H.Wang, J.R. Lee, J. of Microscopy, Vol.191, pp275 (1998); H. Wang, Y.M.Chiang,
J.Am.Ceram.Soc. Vol.81(1), pp89 (1998) and references therein
[Cla 1999]
[Cor 1990]
[CRI 1976]
Lightning Protection Design Standard Committee, Lightning Protection Design Guidebook for
Transmission Lines, CRIEPI Report, No.175031 (in Japanese), 1976.
[CRI 1989]
Lightning Protection WG: Winter lightning performance on the coast of the Sea of Japan, CRIEPI
Report, No.T10 (in Japanese), 1989
[CRI 1995]
Subcommittee for Power Stations and Substations, Lightning Protection Design Committee,
Guide to Lightning Protection Design of Power Stations, Substations and Underground
Transmission Lines, CRIEPI Report, No.T40 (in Japanese), 1995.
[CRI 2003]
[Dar 1998]
Darveniza, Saha, Surface flashover on metal-oxide varistors blocks, IEEE international conference
on conduction and breakdown in solid dielectrics, June 22-25, Vsters, Sweden, 1998
[Den 1998]
[DLT 2002]
DL/T 815-2002: Polymeric housed metal oxide surge arresters for a.c. power transmission lines,
Issued on April 27, 2002, Implemented on Sept.1, 2002, Issued by State Economy and Trade
Commission of P.R. China
[Dor 1985]
[Eda 1984]
Eda, Destruction mechanism of ZnO varistors due to high currents, Journal of applied Physics,
Volume 56, Number 10, Page 2948-2955, 1984
[Elf 2002]
M.Elfwing, E.Olsson, Uppsala Dissertations, Faculty of Science and Techn., 686 (2002), ISBN 91554-5236-1.
[EN 2012]
EN 50526-1:2012, Railway aplications-Fixed installations- D.C. surge arresters and voltage limiting
devices-Part 1: Surge arresters.
[Fan 1993]
J.Fan, R.Freer, J.Mater.Sci., Vol.28, pp1391 (1993); J.Am.Ceram.Soc., Vol.77(10), pp2663 (1994);
J.Appl.Phys., Vol.77, pp4795 (1995)
[Fuk 1997]
Analytical study of failures rates of surge arresters on power distribution lines due to winter
lightning. Japan Korea joint symposium on ED and HVE October 1997 Fukuoka Japan.
Page 142
[Fuk 2012]
T. Fukano, M. Mizutani, Y. Kayano, Y. Kasuga, and H. Andoh, Development of GIS type Surge
Arrester applying Ultra High Voltage Gradient ZnO Element, Transmission and Distribution
Conference and Exposition (T&D), 2012 IEEE PES.
[GB 2000]
GB 11032-2000: Metal-oxide surge arresters without gaps for a.c. systems, Issue date: January 3,
2000, Implementation date: August 1, 2000, Issued by China State Bureau of Quality and
Technical Supervision
[Glo 1981]
A.B.Glot, pp194 in [Lev 1989], A.B.Glot, S.V.Firsin, A.Y.Yakunin, Izvestiya vuzov USSR, Physics,
Vol.5, pp101 (1981)
[Gh 2006]
[Gre 1984]
[Gre 1986]
F.Greuter, G.Blatter, M.Rossinelli, F.Schmckle, Mater.Sci.Forum, Vol. 10-12, pp235 (1986) and in
[Lev 1989]
[Gre 1989]
F.Greuter, R.S. Perkins, M.Rossinelli, F.Schmckle, ABB Review, Vol.1, pp1 (1989)
[Gre 1990]
[Gre 1995]
[Gre 1998]
[Gup 1990]
[Gut 2003]
[Hac 2008]
K. Hachiya, H. Hyodo, N. Honjo, Lightning Strike Aspect for Wind Turbines in Winter Season
Observation Results at Nikaho Wind Park in Japan , 29th ICLP, Uppsala, 9a-4, 2008.
[Hag 1997]
[Ham 1992]
Hamel, St. Jean, Metal Oxide Surge Arresters for EHV Systems - A Comparison Between the ANSI,
IEC and ACNOR Standards, Cigr 1992 Session, Aug. 30 Sept. 5, Report 33-201)
[He 2004]
see e.g. compilation in: J.He, R.Zeng, Q.Chen, S.Chen, Z.Guan; IEEE Trans. Power Delivery,
Vol.19(1), pp138 (2004)
[He 2007]
He, Hu, Discussions on nonuniformity of energy absorption capabilities of ZnO varistors, IEEE
Transactions on power delivery, Volume 22, Number 3, 2007
[Hei 2001]
[Hen 1989]
[Hin 1987]
Page 143
[Hin 1989]
Hinrichsen, Peiser, Simulation of the electrical and thermal behaviour of metal oxide surge
arresters under ac-stress, 6th ISH, New Orleans, 28.8.-1.9.1989, Paper 26.04
[Hin 1990]
[Hin 1994]
V. Hinrichsen, H. Fien, H.-B. Solbach, J. Priebe. Metal Oxide Surge Arresters with Composite
Hollow Insulators for High Voltage Systems; CIGR 1994 Session, 28 August - 3 September 1994,
Paris, paper 33-203
[Hin 2003]
Volker Hinrichsen, Latest Designs and Service Experience with Station-Class Polymer Housed
Surge Arresters. World Conference on Insulators, Arresters & Bushings. Marbella (Mlaga), Spain,
November 16-19, 2003. Proceedings pp. 85-96
[Hin 2007]
Hinrichsen, Reinhard, Richter (on behalf of Cigr WG A3.17), Energy handling capability of highvoltage metal-oxide surge arresters Part 1: A critical review of the standards, Cigr SC A3
Technical Colloquium, Rio de Janeiro, September 12/13, 2007
[Hin 2008]
Hinrichsen, Ghler, Clemens, Steinmetz, Riffon, External Grading Systems for UHV Metal-Oxide
Surge Arresters - A New Approach to Numerical Simulation and Dielectric Testing, Cigr 2008
Session, 25. 29. August 2008, Paris, Report A3-205
[Hin 2009]
Hinrichsen, Reinhard, Tuczek, Recent Experimental Findings on the Energy Handling Capability of
Metal-Oxide Varistors for High-Voltage Applications, 2009 World Congress on Insulators, Arresters
and Bushings, Crete, May 10-13, 2009
[IEC 1991]
IEC 60507 Ed.2, 1991-04: Artificial pollution tests on high-voltage insulators to be used on a.c.
systems
[IEC 2000]
IEC 60099-5, Ed. 1.1, 2000-03: Surge Arresters - Part 5: Selection and application
recommendations
[IEC 2002]
IEC 60099-6, First Edition, 2002-08: Surge arresters Part 6: Surge ar-resters containing both
series and paral-lel gapped structures Rated 52 kV and less
[IEC 2006]
IEC 60099-4, Ed.2.1, 2006-07: Surge arresters Part4 Metal-Oxide surge arresters without gaps
for a.c. systems
[IEC 2009]
IEC 60099-4, Edition 2.2, 2009-05: Surge arresters - Part 4: Metal-oxide surge arresters without
gaps for a.c. systems
[IEC 37]
IEC document 37/317/CDV, Amendment 1-f 3 to IEC 60099-4 Ed. 2.0 Surge Arresters - Part 4,
Metal-oxide surge arresters without gaps for a.c. systems, Clause 8.7: Short-circuit tests
[IEE 2005]
IEEE C62.11: IEEE Standard for Metal-Oxide Surge Arresters for AC Power Circuits (>1 kV), 2005
[IEE 2006]
IEEE 693-2005, IEEE Recommended Practice for Seismic Design of Substations, 2006-5.
[Ike 1981]
G.. Ikeda, S. Sumi, Lightning Parameter in Japan, Research Letters on Atmospheric Electricity,
Vol.1, pp. 41-44 (in Japanese), 1981
[INMR 2002]
Arresters: Market Forces, Current Technologies & Future Directions Part 1 of 2 Distribution
Arresters INMR; November/December 2002, Vol-ume 10, Number 6
[INMR 2003]
Arresters: Market Forces, Current Technologies & Future Directions Part 1 of 2 HV Arresters
INMR; January/February 2003, Volume 11, Num-ber 1, pp. 22-44
[Ima 1984]
M. lmataki, K. Ujita, Y. Fujiwara, S. lshibe, and T. Nitta, Advanced metal oxide surge arrester for
gas insulated switchgear GIS , IEEE Trans. Power App. Syst., Vol. PAS-103, No. 10, 1984.
Page 144
[Ish 2008]
[Ish 2004]
[JBT 2005]
JB/T 8952-2005: Polymer-housed metal oxide surge arresters without gaps for a.c. systems (IEC
60099-4: 2001, NEQ), Issued on Mar. 19, 2005 Implemented on Sept. 1, 2005, Issued by State
Development and Reform Commission of P.R. China
[JEA 1998]
[JEC 2003]
JEC-2371-2003: Standard of the Japanese Electrotechnical Committee for Insulator Housed Surge
Arresters
[JEC 2003]
[Kan 1997]
Kannus, K., Lahti, K., & Nousiainen, K., Aspects of the performance of metal oxide surge arresters
in different environmental conditions, CIRED 97, Birmingham, UK, 2-5 June, 1997, paper 1.4.1
[Kan 1998]
Kannus K., Lahti K. & Nousiainen K., AC and Switching Impulse Performance of an Ice-Covered
Metal Oxide Surge Arrester, IEEE Transactions on Power Delivery, Vol. 13, No. 4, October 1998.
pp. 1168 1173.
[Kle 2004]
[Kno 1985]
[Kno 1986]
[Kob 1998]
[Koh 2000]
A.F.Kohan, G.Ceder, D.Morgan, Ch.G.Van de Walle, Phys.Rev.B, Vol. 61(22), pp15019 (2000)
[Lah 1999]
Lahti K., Kannus K. & Nousiainen K.,A Comparison Between the DC-Leakage Currents of Polymer
Housed Metal Oxide Surge Arresters in Very Humid Ambient Conditions and in Water Immersion
Tests, (IEEE TrPD, Vol. 14, No. 1, January 1999. pp. 163-168)
[Lah 2002]
Lahti K., Kannus K., Nousiainen K., Behaviour of Internal Leakage Currents of Polymer Housed
Surge Arresters suffering internal moisture problems, (INSUCON Conference, Berlin, Germany,
2002, pp. 105 111)
[Lah 2003]
Lahti K.,Effects of Internal Moisture on the Durability and Electrical Behaviour of Polymer Housed
Metal Oxide Surge Arresters, (Tampere University of Technology, Publications 437, 2003.
91p.+Appendixes)
[Lat 1983]
Lat, Thermal properties of metal oxide surge arresters, IEEE Transactions on power apparatus and
systems, Volume PAS-102, Number 7, Page 2194-2202, 1983
[Lat 1985]
Lat, Analytical method for performance prediction of metal oxide surge arresters, IEEE
Transactions on power apparatus and systems, Volume PAS-104, Number 10, Page 2665-2674,
1985
[Lee 1999]
Y.J.Lee, H.D.Hwang, S.W.Han, H.B.Gang, Trans. Korean Inst.of Electrical Engineers C, Vol.48(2),
pp109 (1999)
Page 145
[Len 2000]
Lengauer, Rubesa, Danzer, Finite element modelling of the electrical impulse induced fracture of a
high voltage varistor, Journal of the European Ceramic Society, Volume 20 (2000), Page 10171021
[Len 2000]
[Lev 1989]
[Mah 1983]
[Mah 2001]
[Mar 1992]
[Mar 1996]
Martinez, Zanetta Jr., Comments on Metal Oxide Surge Arresters Surges Energy Absorption
Capacity, Conference Record of the 1996 International Symposium on Electrical Insulation,
Montral, Qubec, Canada, June 16-19, 1996, pp. 498...501
[Met 2007]
[Mic 2007]
[Miy 1992]
K. Miyake, T. Suzuki, K. Shinjou, Characteristics of Winter Lightning current on Japan Sea Cost
IEEE Trans., PWRD, Vol. 7, pp. 1450-1456, 1992.
[Miz 1983]
A.Mizukoshi, J.Ozawa, S.Shirakawa, K.Nakano, IEEE Trans. Power Appar.Syst., Vol.102(5), pp1384
(1983)
[Nai 1996]
[Nie 1989]
[Oka 1986]
[Ost 1992]
Osterhout, Comparison of IEC and U.S. Standards for Metal Oxide Surge Arresters, IEEE
Transactions on Power Delivery, Vol. 7, No. 4, October 1992, pp. 2002-2011
[Per 2003]
R. S. Perkins: The Impact of Metal Oxide Disk Performance on Surge Arrester Operation. Proc.
World Conf. & Exhibition on Insulators, Arresters & Bushings, Marbella, Spain, Nov.16-19, pp193
(2003)
[Per 2005]
R. S. Perkins, F. Greuter, V. Hinrichsen and B. Richter. Integrated surge arrester systems. CIGR SC
A3&B3 Joint Colloquium, Tokyo 26-27 September 2005. Paper 201
[Pik 1984]
G.E.Pike, Mater.Res.Soc.Symp.Proc., Vol.5, pp369 (1982); Phys.Rev.B, Vol.30,pp795 & 3274 (1984)
[Pik 1985]
[Pry 1998]
B.M. Pryor, B. Richter, Overvoltage protection in open air terminal and GIS in the 145 kV
distribution system, Trends in Distribution Switchgear, 10-12 November 1998, Conference
Publication No.459, IEE 1998.
Advances in Varistor
Page 146
[Rei 2008a]
Reinhard, Hinrichsen, Richter, Greuter (on behalf of Cigr WG A3.17), Energy Handling Capability
of High-Voltage Metal-Oxide Surge Arresters Part 2: Results of a Research Test Program, Cigr
2008 Session, Paris, August 24-29, 2008, Report A3-309
[Rei 2008b]
[Ric 2007]
B.Richter, M. de Nigris, V.Hinrichsen: MO surge arresters for voltage systems above 550 kV Experience and challenges for the future, IEC/CIGRE UHV Symposium, Beijing, July 2007, Report
2-5-1
[Ric B 2007]
B. Richter: A critical review of the actual standard IEC 60099-4: Metal-oxide surge arresters
without gaps for a.c. systems Cigre International Technical Colloquium Rio de Janeiro 2007
[Rin 1997]
Ringler, Kirkby, Erven, Lat, Malkiewicz, The energy absorption capability and time-to-failure of
varistors used in station-class metal-oxide surge arresters, IEEE Transactions on power delivery,
Vol. 12, No.1, 1997
[Rin 1997]
K.G.Ringler, P.Kirkby, C.C.Erven, M.V.Lat, T.A.Malkiewicz, IEEE Trans. Power Delivery, Vol.12(1),
pp203 (1997)
[Sak 1989]
[Sat 2007]
[Sch 1992]
W. Schmidt, B. Richter, G. Schett, Metal Oxide Surge Arresters for Gas-insulated Substations (GIS)
Design Requirements and Applications, Cigr Paper 33-203, 1992 Session, 30 August-5
September, Paris.
[Sch 1996]
Walter Schmidt. New POLIM medium-voltage surge arresters with silicone insulation;
ABB Review 2-96
[Shi 1997]
[Sky 2002]
Torbjrn Skytt, Hans E. G. Gleimar. Changing of the guard Polymer re-places porce-lain for surge
arresters; ABB Review 1/2002, pp. 43-47
[Ste 2003]
K. Steinfeld, B. Kruska, W. Welsch. Manufacturing and Application of Cage De-sign High Voltage
Metal Oxide Surge Ar-resters; XIIIth International Symposium on high Volt-age En-gineering (ISH),
Delft/Ne-therlands, August 25 to 29, 2003; Proceedings: Millpress, Rotterdam, ISBN 90-77017-798
[Ste 2004]
L.Stenstrm, R.Osterlund et.al., Cigre 2004, A3, PS1, Question Q.1-6 (2004)
[StJ 1990]
St-Jean, Petit, Latour, Determination of metal-oxide arrester operating limits by a temperaturemargin concept, IEEE Transactions on Power Delivery, Vol. 5, No. 2,. April 1990
[Str 1995]
R.Strmpler, P.Kluge-Weiss, F.Greuter, Adv. in Sci. and Techn., Vol.10, pp15 (1995); F.Greuter
et.al., J.Electroceram., Vol. 13, pp739 (2004); H.J.Gramespacher et.al., Proc. 2003 Intern.Conf.
insul. power cables A.6.3
[Str 2001]
R.Strobl, W.Haverkamp, G.Malin, F.Fitzgerald, 2001 IEEE/PES Transm. and Distribution Conf.,
Vol.2, pp771 (2001); U.Amerphol, M.Kirchner, Boettcher, G.Malin, 2002 CIGRE session, no 21-106
[Stu 1990]
F.Greuter,
[Suk 1988]
[Tec 2006]
Technical Report IEC 62271-300 1st edition, High-voltage switchgear and control gear Part 300:
Seismic qualification of alternating current circuit-breakers, 2006-11.
[Tra 1994]
TransiNor Technical Report no 9.405 1994 Energy capacity and short circuit safety of metal
oxide surge arresters A. Schei, V. Larsen, O. Kvien.
[Tua 1988]
[Tuc 2009]
[Uma ]
[Uma 1987]
[Ver 1992]
[Voj 1996]
A.Vojta, Q.W.Wen, D.R.Clarke, Comput.Mater.Sci, Vol.6, pp51 (1996) and J.Appl.Phys, Vol.81,
pp985 (1997) and J.Am.Ceram.Soc., Vol.80, pp2086 (1997); C.W.Nan, D.R.Clarke, J.Am.Ceram.Soc.,
Vol.79, pp3185 (1996) and Q.Wen, D.R.Clarke, Ceram.Transactions, Vol.41, pp217 (1994)
[Voj 1997]
[Wan 1998]
[Zha 2005]
Page 148
Annual
Report
Conf.
on
Electr.