Harmonic filter design for IEC 61000 compliance

Marius JansenALSTOM Grid

Power System CompensationBrisbane, Australia

marius.jansen@alstom.com

Abstract

The paper1 provides a guideline for the selection ofharmonic filters to ensure that compliance to har-monic emission limits is achieved. Three aspectsof the process are highlighted:

1. Gathering harmonic information from exist-ing networks and applying the information inthe model,

2. Selecting a harmonic filter structure to ensurethat compliance is ensured under a variety ofsituations, and

3. Designing the various filter components to en-sure safe and reliable operation under all fore-seeable network conditions.

1 Introduction

The need for harmonic filter design in power sys-tems is clear: changing loads, capital expenditure-optimised networks and the trend towards dis-tributed generation requires that reactive powercompensation devices often have to double up asharmonic mitigation solutions. While power qual-ity standards are now widely known and applied— with the associated proliferation of power qual-ity solutions — the application of these solutions isnot always well understood. Good harmonic filterdesign can appear to be a black art practised by aselect few engineers.

A substantial body of knowledge has in fact beenaccrued over the years: for example a very good

1Presented at the DIgSILENT User Conference, February2011, Melbourne, Australia

guide can be found in a CIGRE publication relatingto harmonic filter design in HVDC applications [1].Careful reading of this and other publications indi-cate that in most cases experience, detailed knowl-edge of networks, filters and loads, and realism inthe selection of components are the guiding princi-ples in harmonic filter design, and in recent yearsthe process has been enhanced by extremely pow-erful network analysis tools. The article combinesa description of the compliance process with thecapabilities of the most powerful network analysistools to describe a structured approach to harmonicfilter design.

The article is presented from the perspective of amedium or high voltage network however the prin-ciples are valid for all voltage levels.

2 Analysis for compliance

The standard for harmonic distortion emission lim-its most commonly referred to in grid connectionagreements is IEC 61000-3-6 (of which AS/NZS61000-3-6 is a very close if not verbatim copy) [2].

Jurisdiction will normally determine which stan-dard applies (The equivalent North American guideis IEEE 519 [3]) but all standards essentially oper-ate under the principle that the amount of harmonicemission tolerated from each load in the network isdetermined by the nature of the network, the natureof the load, and the vulnerability of the network toharmonic distortion.

The emphasis is on permissible voltage distor-tion in the IEC 61000 standard, the principle be-ing that voltage distortion is seen by all loads con-nected to a node and therefore compliance results

Page 1

in safe operation of all connected equipment.

Figure 1: Indicative planning levels (extract from[2], Table 2)

One of the most commonly quoted tables fromIEC 61000-3-6 is reproduced in figure 1. Itpresents indicative values of planning levels forharmonic voltage (in percent of the nominal volt-age) in MV, HV and EHV power systems. Unfortu-nately the table is sometims incorrectly interpretedas being the final word on what power systems cantolerate. Rather, network operators are responsiblefor setting planning limits for their own networksgiven immunity levels of the equipment in theirnetwork, and must then calculate emission limitsaccording to the procedures in the standard for ev-ery new load to ensure their own planning levelsare not exceeded. The table is only a guideline,and the standard deals in detail with the process to-wards setting these emission limits.

The derived emission limits are the compliancelimits. They are always lower than the planninglimits. Emission limits are generally stated as volt-age distortion at a busbar in the absence of otherloads or background distortion.

The process of determining emission limits isnow implemented at most transmission utilities inAustralia, with gradual penetration in distributionand occasionally in large industrial applications.

2.1 Compliance process

The purpose of filter design is to provide a designthat ensures reliable and safe operation with quan-

tified sensitivity to component parameter changes,frequency variations and network changes. A con-ceptual process flow is shown in figure 2.

Figure 2: Compliance process

In the first stages of analysis, the focus is onthe model that is being constructed — informa-tion about the loads that are causing problems, thedetails of the network and the possible filters thatmay be implemented. Once a network model isconstructed with confidence, filter solutions canbe postulated and the resulting network tested forcompliance.

During compliance testing, the focus is on en-suring that the solution will meet requirementsunder all possible operating conditions. Operat-ing conditions can vary as a result of networkchanges such a substation augmentation, switchingor changed generation conditions, and are gener-ally presented by the network planning authorityas a set of impedance curves at one or more bus-bars in the network. New loads may be added, orexisting loads removed or changed.

The proposed solution is also subject to variation— changes in temperature or component drift ortolerance can change the behaviour of the filter ina material way.

Whatever the number or type of these possibleoperating conditions, it is important that they are

Page 2

quantified and described so that it is clear what thescope of any compliance statement is, and so thatany possible non-compliance in future can be ref-erenced to the scope of the study.

Once compliance has been verified, the empha-sis moves to design, where reliability and costs be-come important. A similar approach is used forthe analysis for compliance and reliability in thesense that all possible operating conditions need tobe considered:

1. Compliance looks at the performance of thenetwork, whereas reliability looks at the fil-ter components. Under compliance testing, alloperating conditions are considered and theproposed solution must limit emissions underall of these.

2. For reliability, the worst operating condition isfound and the design ensures that equipmentis rated appropriately.

2.1.1 Loads

Modern network analysis tools make modeling ofnetworks a relatively simple task. Detailed, com-plex models of networks and loads can be con-structed from scratch, or shared across hardwareplatforms and translated between software pack-ages with relative ease. It can therefore be a trap toassume that the software will do the critically im-portant tasks for you as well, such as ensuring thatthe loads that contribute to the problem are mod-eled accurately.

It is highly recommended if at all possible thatmeasurements of existing loads and voltage distor-tion are recorded, at least at the point of connec-tion of the filter. This has two real benefits: themeasured values can be used to calibrate the model(where the model is used to predict for examplethe harmonic voltage spectrum at a certain bus,which is then compared to the measured values)and it will allow the designer to become familiarwith the site conditions such as space, environmen-tal conditions, access, and preferences of operatingpersonnel. Load measurements can of course bescaled for load growth or changes, and should bedone with consideration the relevant standard [4]

for such work, so that future measurements can berepeated on the same basis.

It is important to take into account the nature ofthe load and the type of work that is going to arisefrom the measurement. For static loads, the situa-tion is quite simple, but presentation of results andfurther modeling of stochastic loads, where har-monic compliance is closely related to probabilitydistributions needs special consideration.

Care should be taken when using current and/orvoltage transformers to measure in medium andhigh voltage networks. It is accepted that lightlyburdened current transformers are quite linear tofairly high frequencies [5], but when the CT bur-den is high the ratio error becomes large, and inany case the phase error at high harmonics makephase measurement unreliable.

Measuring harmonics in high voltage power sys-tems with inductive and in particular capacitivevoltage transformers may cause large errors. Forexample, tests carried in Norway [6] revealed er-rors from 80% to +1200% (0.2-12 pu) of correctvalues.

Inductive voltage transformers can have a rela-tively small error up to 29th harmonic (1450 Hz)while capacitive voltage transformers had large er-rors at a few hundred Hertz. It is theoretically pos-sible, but practically very difficult to calibrate highvoltage instrument transformers, even if such cali-bration may correct the device phase and ratio erroracross the measurement frequency range.

Inductive voltage transformers are preferred forharmonic measurements wherever possible, andmeasurements that purport to indicate direction ofharmonic power flow should be treated with greatcare specifically due to the large phase errors.

2.1.2 Network

The process of creating a network is mostly me-chanical. Re-using existing models can save timeand provide additional confidence in the validity ofthe model. It is important that all the topology andnetwork variations are built into the model from theoutset. Topology variations can relate to changes infault level, loading, generation, connected lines andplanned outages, while network variations includefactors such as voltage deviations, unbalance, and

Page 3

frequency deviations.Once these variations are documented and

agreed, the ability of the model to produce the nec-essary data for each scenario needs to be tested.It is a good idea for quality management purposesthat a record is kept of the output data validitycheck.

The harmonic analysis tool will in most casesneed to be able to

1. perform unbalanced harmonic load flow,

2. analyze inter- and subharmonics, and

3. modeling the frequency dependent character-istics of all network elements.

It is important to calibrate the model — againstmeasurements if at all possible, or against existingstudies or known information such as fault level ornormal load flow operations — before the real anal-ysis begins.

2.1.3 Filters

It is useful to define some general guidelines thatmust be kept in mind in the selection and applica-tion of harmonic filters:

1. Select a filter configuration that is feasible interms of design and impact on the network,

2. Obtain as realistic as possible model for com-ponents: resistance of reactors, manufacturingtolerances of components, sensitivity to aging,temperature,

3. Document the range of variations in induc-tance, capacitance and resistance, and

4. Keep it as simple as possible: minimum steps,minimum components.

A harmonic filter can be simply defined as adevice or combination of devices that is intendedto reduce the potentially damaging effects of har-monic voltage distortion (such as increased cost,network losses or damage to equipment) caused bynon-linear loads in the network.

A harmonic filter can contain a single branch or acombination of several, it can be located at a single

node or be distributed across several nodes, and itcan be be active or passive. It is also possible thata harmonic mitigation solution is achieved as partof equipment selection, for example in transformerimpedance or vector arrangement.

Figure 3: Principle of current divider

A passive filter works as a harmonic filter be-cause it presents a lower impedance path for har-monic current at a particular frequency than the restof the system, as shown in figure 3. The current IL

is distorted by the non-linear load. As a result ofthe current divider consisting of the network andthe harmonic filter, with impedance Zn and Zf re-spectively, some network current is diverted intothe filter according to In = Zf/(Zn+Zf )×IL, andthe resulting current In flowing into the networkresults in harmonic voltage distortion of Vn =In × Zn.

The filter impedance can therefore be designedto reduce network voltage distortion. The amountof harmonic absorption is a function only of thefilter configuration and the network impedance.

The complexity of the task becomes appar-ent when reviewed against a typical frequency-dependent impedance of a transmission system,such as shown in figure 4. A specific function ofharmonic filter impedance must be designed in or-der to present a relatively low impedance path for

Page 4

harmonic current at frequencies where harmonicabsorption is desired, under all known and possi-ble network conditions.

2500.02002.01504.01006.0508.0010.000 [Hz]0.1

1

10

100

1000

10000

41520_GDNSTH1D: Network Impedance, Magnitude in Ohm

Z

Date: 1/18/2011

Annex: /2

DIg

SILE

NT

Figure 4: Typical transmission system impedance

The selection of harmonic filter type is deter-mined by many factors such as the allowable reac-tive power rating (which is determined by accept-able voltage variations when the bank is switchedin and limitations on power factor at the pointof connection), loss evaluation criteria, availablespace on site, budget and availability and reliabilityrequirements.

For passive harmonic filters there are a limitednumber of building blocks – filters are constructedout of combinations of capacitance, inductance andsometimes resistance. The performance of the filteris determined by the rating and topology of thesesimple building blocks. The topologies for some ofthe more common filter types are shown in figure 5.The main categories used to describe types (tuned,damping and order) are illustrated in the followingdescription.

Single tuned This is the simplest filter topology,consisting of a reactor connected in serieswith the capacitor bank. By choosing the ca-pacitance and inductance to achieve a seriesresonance at one specific harmonic order, avery low impedance path, limited only by theresistance in the reactor, is created for one har-monic current. The advantages of this con-figuration are simplicity, cost, low losses, andunless extremely sensitive tuning is required,stable operation that is not significantly influ-

Figure 5: Filter topologies: (a) Single tuned (b) Ctype (c) Damped single tuned (d) Double tuned

enced by component variations or tempera-ture.

C type In this topology an auxiliary capacitor isconnected in series with the reactor and istuned to form a fundamental frequency bypassof the resistor. By creating a tuned filter sec-tion C2-L within a 2nd order filter, virtuallyall fundamental current is excluded from theresistor. At frequencies above fundamental,harmonic current flows through R achievingthe desired damping.

Damped single tuned In this filter topology adamping resistor is connected in parallel withthe series reactor. The presence of the resis-tor broadens out the frequency response ofthe filter which introduces two beneficial ef-fects: The filter is now less sensitive to thede-tuning effects of frequency drift, ambienttemperature variation and component toler-ance effects, and the filter response can covera number of harmonics, therefore it may bepossible for example that two harmonic orderscan be addressed with a damped single tunedfilter. By adding the resistor, filter losses havebeen increased both at harmonic frequencieswhere it is needed and at fundamental fre-quency where it is not. These higher lossescan be prohibitive.

Double (and more) tuned This type of filter issubstantially equivalent to two (or more) par-allel connected tuned filters but is imple-

Page 5

mented as a single combined filter. The re-active power rating of the combined doubletuned filter would be the sum of the ratingsof the two tuned filters. By combining twotuned filters virtually any Mvar split can beaccommodated between the lower and upperfrequency components.

This allows the possibility of incorporat-ing a small Mvar rated filter, which on itsown would be an uneconomic design, into alarger filter to form a double tuned filter. Iftwo single-tuned branches were used instead,there could be a minimum filter size problemdue to connecting a possibly very low Mvarrated filter (as required for one of the two fre-quencies) on to an HV busbar. This problemcan be overcome in most cases by the use ofdouble-tuning. Another advantage is that thedouble tuned filter has high impedance at afrequency between the two (low impedance)single tuned frequencies, and this frequencycan be selected by changing the filter compo-nent values. It is therefore possible to createa blocking filter, for example for audio fre-quency load control systems, by using such afilter combination.

2.2 Compliance test

The compliance test compares the stated emissionlimits of a load or installation as mitigated by aspecific harmonic filter solution, with all the pos-sible variations that have been identified as possi-ble and realistic for the network and the harmonicfilter. With N possible topologies, K network pa-rameters and M filter parameters that can change,N ×K ×M can easily result in thousands of dis-crete scenarios that require evaluation. Commonsense can reduce the number of scenarios by elim-inating impossible combinations, but the result setwill still be very substantial.

This is a task that requires detailed planning be-fore any analysis is done. Analysis software likeDIgSILENT PowerFactory can be automated toread in complete sets of parameters of networks,variations in network parameters, and different fil-ter parameters, and can produce the required testsextremely efficiently. In fact, such an automated

Figure 6: Proliferation of scenarios

process is essential if an exhaustive analysis of any-thing more than the most elementary networks.

The analysis process typically produces tagged(each result is tagged with the parameters of thescenario that it relates to) sets of harmonic spectraat multiple busbars depending on the scope of thestudy.

The analysis software can perform all the neces-sary tests and simply produce a Pass/Fail result butfor reporting purposes a clear and compact presen-tation format is required. This format must not be asubset of the results showing some selected typicaloutcomes as the selection process can be biased,and it must also not be a comprehensive replica ofthe results at all the busbars, for all scenarios underreview, as this will result in information overload.

A simple tool used in statistical analysis for sum-marizing data is proposed as a solution. A sensible,compact approach to data representation is the boxand whisker plot.

In figure 7 the compliance level is presented as asolid background bar of blue. For each harmonicorder, the results from all the simulated scenar-ios are presented in summary form as a red linespanning from the maximum distortion level to theminimum distortion level. A rectangle shows thespread of values between the 10th and 90th per-centile. This type of graph summarizes informa-tion about compliance to emission limits and sen-sitivity of the solution to network changes that nor-mally occupy large swathes of pages in harmoniccompliance reports.

As an example from this graph, it is clear thatemission limits for all but the second harmonic

Page 6

are achieved, and that the solution being evaluateddoes not achieve compliance at the second har-monic under any circumstances. The large varia-tion in 13th harmonic distortion indicates that thesolution is sensitive to network changes, and mayalso require further review.

Figure 7: Optimized data representation

With this data representation, the necessary ad-justments can be made to filter parameters basedon performance in specific scenarios, and the com-pliance test repeated. Compliance can therefore bereached with an optimal cost solution in terms ofcapital and lifetime costs.

3 Design for reliability

A very similar approach is followed to determinethe required component rating. By identifying theworst case voltage stress, fundamental or harmoniccurrent levels in each of the components in the net-work, the exact safe minimum ratings can be deter-mined with ease.

The various components are rated and specifiedaccording to the applicable standards, such as IEC60871 for capacitor units and AS/NZS 1028 (orIEC 60076-6) for reactors.

Care should be taken in applying the standards,for example the capacitor standard requires that thevoltage rating of capacitor units should be equal tothe arithmetic sum of all the voltages (fundamentaland harmonic) that may continuously be applied to

the capacitor. If a load is known to be erratic ordoes not operate for long periods, a lower continu-ous voltage rating may be required than if the loadwas constant.

Note also a discrepancy that has recently arisenbetween AS 1028 and IEC 60076-6. For harmonicfilter reactors, AS 1028 states that the required cur-rent rating must be equal to the RMS current actu-ally expected to flow in the reactor. In IEC 60076,harmonic filter reactors are required to have a cur-rent rating of 1.43× IN , the nominal current of thebank, frequently resulting in a much higher currentrating (and cost) than if AS 1028 was applied.

Other issues relating to design but not exploredfurther here are those of acceptable sound emis-sions, seismic and wind loading, and the impor-tance of standardisation.

A description of the various additional aspectsof filter design and specification can be found insection 11–15 of the CIGRE guide [1]. The mostimportant considerations are described below.

3.1 Transient behaviour

It is of specific importance to review the transientbehaviour of the installation — the stresses on in-dividual components during energisation of the in-stallation or nearby equipment may have an impor-tant bearing on the short time and impulse with-stand ratings of the installation and all the com-ponents. Energisation (including de-energization)studies must be carried out, taking into accountthe nature of point on wave switchgear that maybe used. Here the challenge will be similar tothat described above for steady state harmonic fil-ter design - to determine with certainty that theworst case condition has been found and quanti-fied. Note that these may arise due to events out-side the filter, such as reclose operations of nearbycircuit breakers, or transformer energization. Theoutcome of this analysis will enable the designerto recommend and rate any required impulse with-stand and/or surge arresters.

3.2 Protection

Protection of harmonic filters take into accountall the normal functions of power system protec-

Page 7

tion. For the bank, and each branch, considerationshould be given to short circuit protection, over-current and earth fault protection, thermal over-load protection,over/under voltage protection, andbusbar- and breaker failure protection.

Components inside the bank can be protectedby means of unbalance protection for capacitorswhich is properly coordinated with the operation ofinternal fuses in the capacitor units. It is also pos-sible with modern protection equipment for indi-vidual components to be protected by special mea-surement functions such as fundamental and har-monic frequency current measurements.

4 Conclusion

The approach to harmonic filter design presentedin this article provides a clear description of theprocess to be followed to ensure that complianceto emission limits is achieved, and is capable ofenumerating the exact range of operating condi-tions and other contingencies this this complianceis valid for.

The same approach also enables the determina-tion of the actual worst case scenario on whichcomponent ratings are determined. The approachenables the rating of equipment to be safely op-timized for all actual expected operating condi-tions, avoiding the most unrealistic and conserva-tive worst case conditions that are used as the basisfor equipment ratings in the absence of such infor-mation.

The result is a quantified compliance test with acost and performance optimized filter solution.

References

[1] Technical brochure prepared by WorkingGroup 14.30, Guide to the specification anddesign evaluation Of AC filters For HVDC sys-tems, CIGRE.

[2] AS/NZS 61000-3-6:2001, Limits — Assess-ment of emission limits for distorting loads inMV and HV power systems, Standards Aus-tralia, 1999.

[3] IEEE 519-1992, IEEE Recommended Prac-tices and Requirements for Harmonic Controlin Electrical Power Systems, IEEE, 1992.

[4] AS/NZS 61000-4-7:1999, General guide onharmonics and interharmonics measurementsand instrumentation, for power supply systemsand equipment connected thereto, StandardsAustralia, 1999.

[5] K. Debnath, “Accuracy of distribution cur-rent transformers under non-sinusoideal exci-tation”, in Australasian Universities PowerEngineering Conference and IEAust ElectricEnergy Conference, Darwin, 26-29 September1999.

[6] H. Seljeseth, E. A. Saethre, T. Ohnstad, andI. Lien, “Voltage transformer frequency re-sponse. measuring harmonics in norwegian300 kv and 132 kv power systems”, in Pro-ceedings of the 8th International Conferenceon Harmonics And Quality of Power, Trond-heim, 1998.

Page 8