Bennett1997.pdf

v

GEC A L S T H O M

T&D

THE EFFECTS OF HARMONIC DISTORTION ON EQUIPMENTS. BY : P C BENNETT

V DARLEY

S M ABBOTT

Introduction

The brief for this portion of the Colloquium is to discuss the effects of harmonics on equipments. We trust you will accept that as this contribution was prepared within a Transformer Engineering Department the illustrations used to show the effects concern transformers. It is a truism that all transformers are affected by harmonics given that any power system is contaminated to some degree by them. What is of interest and concern are the extremes, albeit known and in most cases predictable extremes. Some loads are known as classical harmonic generators so far as transformers are concerned. Any transformer connected to a Converter load or an Electric Arc Furnace is by definition connected to a harmonic generator. In the case of the Converter harmonic levels maybe variable, predictable and even controllable. In the case of an Arc Furnace the uncontrolled nature of the arcs during certain times of the melting process produces wholly unpredictable and randomly generated harmonic currents. In order to demonstrate the effects of harmonics on transformers two illustrations are proposed. Firstly a case study of what happens when the effects, very regrettably, are not predicted. Secondly how, with modern standards and computer aids, the effects can be predicted, incorporated into the design and demonstrated by appropriate thermal tests.

CASE STUDY. EFFECT OF HARMONICS The case study concerns a 12 pulse full wave 30kA Rectifier transformer with a rating of approximately 30 MVA.. The unit was manufactured in the early 1970's. One of the initial Design decisions was how to achieve the required 12 pulse operation from a single transformer equipment. Two methods are possible and have been used to build operational units. One method uses two individual transformer core and windings housed in a single tank. One transformer will have its low voltage winding delta connected the other unit will be star connected. These 30 degree shifted outputs are each connected to 6 pulse full wave Converters which are in turn parallelled. The alternative is to assemble and axially displace both sets of windings onto a single 3-limb core. There are a number of economical

511

0 1997 The Institutionof Electrical Engineers. Printed and published by the IEE, Savoy Place, London WC2R OBL, UK.

and physical reasons, not of immediate interest, that make this the preferred alternative and the one concerned in the case study The connections of the main unit are shown in Diagraml. Some details are missing. The regulator scheme that varies the main unit input voltage and any phase shifts to the primary windings that multiply the individual unit 12 pulse output to that required for the complete installation. Some details are shown. The half wave flux resetting saturable reactors used for fine voltage control because, in the example, these were mounted with the main unit.

By convention with power transformers the outermost and hence visible windings are those associated with the HV circuits. With heavy current Converter units this convention is reversed so the outer windings are those of the LV circuit which, for our case study is convenient, given that the problem is associated with these windings. The photograph (Diagram2) shows the original transformer with the transductor assembly mounted above the main unit. In the initial design 16 parallel connected disc coils formed the LV star winding assembled in the top halve of the core whilst 20 parallel disc coils formed the LV delta in the bottom portion of the core. The question was how do these parallel coils share the fundamental and harmonic currents? It was known that the fundamental component of the load current did not share equally between the discs. Prior to computer programs the distribution formulae had been established from tests on models. It was thought that the harmonic components of the load current, of which the 5'h, 7'h, 1 1th and 13'h were the most significant, would share in a similar manner. Unfortunately this was not so. The table of test results (Diagram3a) shows that the pair of disc coils adjacent to the central gap between star and delta carry a disproportional share of the harmonic current. The result of this sharing was a significantly increased rate of life usage of the paper insulation on these coils. Improvement was in two stages. In the first balancing reactors were introduced in series with the central coils. The improvement is shown on in the expanded table of test results. (Diagram3b) Secondly field plotting computer programs were developed that permitted a full analysis of both the fundamental and harmonic current sharing for any winding geometry. This resulted in replacing the disc winding with helical type coils. The further improvement in current distribution when added to the test result table (Diagram3c) shows the original and final current sharing. The photograph (Diagram4) of the new arrangement is not visibly much different from the original, however as the diagram (DiagramS) illustrates not only does this arrangement improve current sharing it significantly lowers the main unit copper losses. This is mainly due to the reduction in eddy current losses especially in the coils adjacent to the central gap.

The motto of this admittedly somewhat old story is to know during the design stage that (a) there are harmonics (b) their magnitude and (c) where they flow within the windings. This requirement is as valid today as it was when the lesson was originally learnt.

512

ALLOWING FOR THE AFFECT OF HARMONICS. We may now assume that the value of the harmonics to which the transformer is to be subjected are known, explore how their affects maybe allowed for at the design stage and how traditional thermal tests maybe modified to demonstrate compliance Firstly another truism, harmonic currents generate heating affects over and above those of the fundamental currents assumed in conventional transformers. To make adequate allowance for the additional heating affects it is necessary to understand what elements within the transformer are affected. The diagram (Diagram6) illustrates the basic temperature rise characteristics of a unit designed to present BS/IEC standards. Three temperature components are guaranteed, two openly and one hidden. The 'Top Oil' temperature rise and the 'Average Winding Rise by Resistance' are guaranteed and thermal tests are conducted to show compliance. The average winding rise is the sum of the Mean Oil rise and a 'Winding Gradient' which is deduced from a resistance cooling curve taken at the end of the thermal test. There is a third important but hidden element namely the 'Hot Spot temperature' within the transformer. As it name implies this is the hottest temperature usually of insulation on a portion of winding conductor and permitted to be 98 C. to present BS/IEC standards. Traditionally this value has been calculated by adding 10%to the 'average ' gradient and adding this to the Top Oil rise and the yearly ambient temperatures. At this point a small digression. Transformers maybe overloaded and the rule governing such operation states, in it's simplest form, that for every 6 C. above 98 C. hot spot temperature the life of a transformer is halved and doubled for every 6 O C. below. In the practical application of transformers users seldom operate units continuously at their nameplate rating, manufacturers seldom achieve the permitted rises (here the Charles Dickens rule applies; winding rise by test 64.5 -result Happiness; winding rise by test 65.5 - result misery) and finally the BS/IEC assumes a 20 C. average ambient temperature which somewhat overstates the case so far as the UK is concerned. The combination of these factor results is a safety margin sufficient to cater for a degree of harmonic contamination that maybe present in a transformer connected to the power system.

To return to the main theme. Harmonics affect each of the three temperature elements referred to above. Firstly they increase the total loss generated and hence affect the oil rise elements and the size of the cooling plant. Secondly they increase the winding eddy losses and so increase the average winding

513

gradient above that obtained solely from the fundamental current. Thirdly the hot spot gradient maybe increased. In this respect studies have shown that the traditional 10% increase of the average gradient to obtain the hot spot value may not be sufficient and a values up to 30% appear more reliable.

A number of standards exist that incorporate calculations for harmonic affects. Initially these were ANSI/NEMA standards but IEC1378 will be the European standard and this will be used to illustrate an actual example. The subject transformer, manufactured and tested recently, has a DC current rating of 45 kAmps.. The photograph (Diagram7) shows the core and windings of the 12 pulse unit with Star/Delta secondaries complete with transdudors in the half wave circuit. The principles on which the new lECl378 Part1 is based are:

(1) the transformer nameplate rating is based on the fundamental components of the voltage and DC current. The usual copper loss and impedance guarantees are based on this rating and tested to IEC76 methods to show compliance. (2) the harmonic current levels to which the transformer is to be subjected shall be agreed. (3) the division between winding eddy loss and the other stray losses shall be assumed to be as the design intention. (4) the winding eddy loss enhancement factor shall vary as the square of the harmonic number. (5) the stray loss enhancement factor shall vary as the 0.8'h power of the harmonic number. In the example the basic fundamental current losses of the tested transformer are shown in the diagram. (Diagram8) Core and Transductor losses have been removed for simplicity. Historically thermal testing with these losses was all that was required and any additional loss enhancement was at the discretion of the manufacturer. The new proposal requires agreed harmonics and the enhancement factors for the eddies and strays for the agreed harmonics of the example are as the diagram. (Diagram9) These factors when applied to the tested losses at fundamental current produce the final loss levels required for thermal tests. The diagram records the harmonically loaded values. (Diagram10) The mean and top oil rises are established using the losses calculated in this manner. The current used to establish the winding gradients during the latter part of the thermal test is a value that would produce the sum of the fundamental current losses plus the harmonic eddy losses of the winding under consideration. The standard therefore provides a method by which harmonically loaded transformers maybe verified by tests prior to being subjected to actual Converter waveshapes,

A word of caution. The approach is crucially dependent upon agreeing a reasonable average level of harmonics. The simple solution of defining harmonics as those of a classical square wave will for many applications result in oversized cooling plant and transformer windings. In the example used the increase in overall losses by using the harmonics of a square wave are shown in the diagram. (Diagram11) Finally lECl378 will ultimately be in two parts. Partl as we have seen is aimed mainly at low voltage Industrial Converter applications whilst Part2 is primarily for HV DC applications. The loss analysis approach is basically the same for both but with an additional copper loss test at higher frequency (150/250 Hz) so that the winding eddy and structural steel stray losses maybe separated with a reasonable degree of accuracy. This approach was not proposed for Partl because some transformers in the scope of the standard will be small in MVA and not all manufacturers have dual frequency test facilities.

hh A a PRIMARY WINDING

SECONDARY WINDING

-0

TRANSFORMER CONNECTION 12 PULSE

3O k A

TRANSDUCTOR

sc BAL

CI

Fundamental

0

1.4 1.5 2.

2.9

CURRENT

WINDING

I

Fundamental

Star Delta

I

Star Delta

I I

%

1.4 1.5

b

2.8

I 1 ,

lf!

1.1 1.1

! 2.2 t

No

0

CURRENT Fundamental

Harmonic

!

ING tar elta Star elta

f

B

11

1

1.4 1.5

1.1 1.1

E

2.8 2.

2.2

E

f

I

I E

i

2.

t

1

1.05 1.05 1.5 1.3

BIAGRA M 4.

5/11

1.o 1.o 17

0.65

20°C YEARLY

I

I

AVERAGE

r HOT SPOT

GRADIENT

5/13

I

5/14

I

12R

EDDIES

HV

40

9

49

LV

60

11

71

BUSBARS

45

STRAYS TOTAL

145

20

STRAYS

TOTAL

10

55

70

70

80

245

T 1.o

5

0.15

0.0225

0.56

7

0.1

0.01

0.5

0.05

0.0025

0.3

0.04

0.0016

0.27 0.12

0.02 0.01

0.0001

I

1.0371

t

!

0.0

S

Gl

4

E-r

0

m

v3 v3

m

m

m

w 00

E-r -

F1

U

?4

el

d-

40

el

4

4

00

m pr: 4

m

*X

m

a Ep

5/17

a a 4

m cn

d4

m

1A

d-

a

m

d-

m & 4 p7

5/18

m m