Harmonic improvement in 12-pulse series-connected line-commutated rectifiers

46

&

www.ietdl.org

Published in IET Power ElectronicsReceived on 19th November 2007Revised on 25th March 2008doi: 10.1049/iet-pel.2007.0014

ISSN 1755-4535

Harmonic improvement in 12-pulseseries-connected line-commutated rectifiersM.E. Villablanca J.I. Nadal F.A. Cruzat W.C. RojasElectrical Engineering Department, University of Santiago, P.O. Box 10233, Santiago, ChileE-mail: [email protected]

Abstract: An innovative method for reducing current distortion on the ac side of 12-pulse series-connected line-commutated ac/dc rectifiers is presented. Furthermore, the overlap conduction of thyristors is completelyeliminated. The methodology involves an accurate shaping of the dc current by using two self-commutatedswitches (IGBTs). This dc-current shaping is reflected back into the shaping of the ac input currents, whichbecome pure sine waves. Both rectifying and inverting operations are possible with a simple control circuit,which is able to deal with both rapid load variations and failures in the self-commutated switches. HVdcapplications of this innovative method are briefly examined, together with experimental results showing boththe steady state and the transient behaviour of the technology by using a 380 V 20 kVA 50 Hz laboratoryprototype.

6

1 IntroductionLine-commutated ac/dc rectifiers have been widely used formany years because of their ruggedness and simplicity.However, they draw distorted input currents with harmfulconsequences over the ac supply system. In recent years,many methods have been proposed in order to reduce thiscurrent distortion.

The present method acknowledges that the shape of the dccurrent defines the shape of the ac current in an ac/dcrectifier. For example, a pure dc current with a negligibleripple component produces a stepped-wave current on theac side. Similarly, the method in this paper shows that atriangular dc current, with a sine-wave outline, definesperfect sinusoidal currents on the ac side. Furthermore, theoverlap angle of thyristors is completely eliminated.

The method has been perfected over the years and the firstproposition was made in the early 1970s [1, 2]. In thatapproach, the input current distortion was reduced byinjecting a harmonic current on the dc side, which was ofdifficult implementation. Later, this problem was solved byusing a thyristor-based auxiliary circuit to shape the dccurrent, and therefore the pulse number of a bridge rectifierwas transformed from 6 to 12 [3].

The Institution of Engineering and Technology 2009

Recent contributions have increased the pulse number of12-pulse series-connected rectifiers. In [4], the pulsenumber was transformed from 12 to 24 and in [5] from 12to 36. This last contribution is shown in Fig. 1, whereby aproper firing of the auxiliary thyristors modifies thewaveform of currents IS1 and IS2 in order to reduce thedistortion of currents IA, IB and IC. Auxiliary thyristors,however, do not open timely at high-level load currents,which is detrimental to the dc-current shaping. In thepresent technology, two self-commutated switches, withoutthis limitation, will shape the dc current through on/offswitching actions.

In the following, Section 2 covers the theoreticalbackground of the technology, and Section 3 theexperimental verification. Also, Section 4 shows HVdcapplications of the technology.

2 Theoretical background of thenew methodFig. 2 shows the proposed rectifier, whereby self-commutatedswitches S1 and S2 accurately shape currents IS1 and IS2 inorder to obtain perfect sinusoidal currents IA, IB and IC.Fig. 3 illustrates how a particular switch current follows itsreference. The switch current is compared with the

IET Power Electron., 2009, Vol. 2, Iss. 4, pp. 466–473doi: 10.1049/iet-pel.2007.0014

IEdo

www.ietdl.org

reference current, and, as the switch current exceeds theupper band in A, the switch is opened and the switchcurrent starts to decay by flowing through the snubbercircuit. When the lower band is crossed in B, the switchis closed and the switch current starts to increase.

Figure 2 Proposed rectifier with high-quality input/outputwaveforms

Figure 1 36-pulse ac/dc rectifier as proposed by Villablancaet al.

T Power Electron., 2009, Vol. 2, Iss. 4, pp. 466–473i: 10.1049/iet-pel.2007.0014

By reducing the hysteresis band, the switch current followsits reference more closely.

Figs. 4 and 5 show the geometry of currents IS1 and IS2.The peak value IM of currents IS1 and IS2 must beproportional to the magnitude of the load current, that is,one magnitude affects the other. In the Appendix, it isdemonstrated that IM ¼ 1.954 IL, where IL is the averageload current.

In the circuit shown in Fig. 2, currents Ia, Ib, Ic, I 0a, I 0b andI 0c are typical ac line currents of a Graetz bridge rectifier(Fig. 6). That is, during one cycle, these currents show apositive area for 1208, zero current for 608, a negative areafor 1208 and zero current for 608. Both areas are modifiedby the shape of currents IS1 and IS2; therefore, currents Ia,Ib, Ic, I 0a, I 0b and I 0c become double-humped.

Fig. 5 shows also two aspects: (1) currents IS1 and IS2 aresynchronised with the beginning of conduction of anythyristor in the respective bridge, and (2) bridge thyristorstransfer their conduction at zero current, eliminating theoverlap angle completely, which will be of interestregarding HVdc applications.

Note that currents I 0a, I 0b and I 0c lag currents Ia, Ib and Ic by308 because of the DYD transformer connection, and there isa turn number difference between both secondaries by afactor

p3.

The equation shown in Fig. 6, namely IA¼ I 0aþ (Ia 2 Ic)/p3, is defined by basic electric laws and transformer concepts.

In the following, it is demonstrated that IA is a pure sine waveby considering windows w1 and w2 in Fig. 6.

Figure 3 Switching process of switches S1 and S2

Figure 4 Sine-wave outline of currents Is1 and Is2

467

& The Institution of Engineering and Technology 2009

468

&

www.ietdl.org

In window w1

IA ¼ Imax sin(wt) (1)

In window w2

IA ¼ 2 � Imax sin(wt � 608)þffiffiffi3p

Imax sin(wt þ 908) (2)

IA ¼ Imax sin(wt) (3)

Switches S1 and S2 through the shaping action becomeperfect current sources and therefore the associated ac anddc current components must have free circulation aroundthe respective ‘current source’ by forming a closed loop.

Figure 6 Theoretical construction of input current IA

Figure 5 Theoretical currents Is1 and Is2


The dc components of currents IS1 and IS2 are identical,whereas the ac components are different (Fig. 5).Therefore, the dc components make a loop through theload and the ac components through capacitors C1 and C2.

Note that unity power factor is reached when using diodes.When using thyristors, however, a certain amount of reactivepower must be injected if unity power factor is desirable.

Fig. 7 shows the control circuit in the block diagram of theproposed ac/dc rectifier. Note that microprocessor 1 (8751)controls the thyristors and provides the system synchronisingsignal to microprocessor 2 (PIC16F873), which in turngenerates two reference currents in digital form for switchesS1 and S2 with the correct frequency and synchronisation.The synchronisation is given by a signal with the starting timeof conduction of any thyristor within the associated bridge.These digital references then enter digital-analogue converters(DAC0808), whereby the average value of the load current IL

is also fed to define the peak value IM of the reference signals(IM ¼ 1.954 IL).

The load current is monitored by a current sensor (LEM).This sensor is based on the Hall-effect and transforms theload current into an equivalent voltage. A low-pass filter(MAX294) then determines the average value of the loadcurrent, which is further amplified by an operationalamplifier (741).

Each reference current is then compared with theassociated switch current (switch currents are monitored bythe same circuit regarding the load current, but without thelow-pass filter). The error between reference and switchcurrents (positive or negative) is defined by aninstrumentation amplifier (AMP03GP). An operationalamplifier (311) digitalises this error, with one representingpositive error and zero negative error. The output of thisoperational amplifier enters microprocessor 3 (PIC16F84),which in turn gives the output signal to control theopening and closing of the associated switch, provided theerror between reference and switch currents is kept within atolerance band.

Figure 7 Control circuit in a block diagram of proposedconfiguration


IEdo

www.ietdl.org

3 Experimental verificationA 380 V 20 kVA 50 Hz converter system was used toimplement the circuit shown in Fig. 2 and experimentalresults are illustrated by using a digital recorder. Themaximum switching frequency was 7 KHz, whereas thehysteresis window was 1.25 A. Villablanca [6–9] shows adetailed description of the implementation regarding bothhardware and software.

In order to demonstrate the efficiency of the technology,the relevant information is shown below when the circuitshown in Fig. 2 is working at semi-industrial scale,together with the identification of switch components

IA ¼ 31:6 A (THDi ¼ 1:37%)V A ¼ 220:1 V (THDv ¼ 1:26%)V L ¼ 965 V IL ¼ 19:8 APT(OUT) ¼ 19:1 kW PT(IN) ¼ 20:8 kWh ¼ 91:8% (h ¼ 95:8%

conventional configuration)

Maximum switching frequency: 7 kHz (Hysteresis window:1.25 A); IGBTs: POWEREX TRANSISTORCM1000HA 24 H; thyristors: CRYDOM EFG16F;snubbers: RADIONICS IGBT CAP 2UF 1 KV.

Efficiency is of paramount importance in industrialequipments, so a lower switching frequency of 1.4 kHz,with a hysteresis window of 3 A was used, resulting inTHDi ¼ 5.5%, but more importantly, an efficiency of94.5% was obtained. Also, the only inductance used in theexperiment was the leakage inductance of the powertransformer with a value of �15 mH.

3.1 Application of an inductive load tothe prototype (R ¼ 47 V and L ¼ 25 mH)

Figs. 8–12 show various current waveforms whenconsidering an inductive load (capacitive loads were alsosuccessfully tested). Note in Fig. 8 the filtering effect of thetransformer leakage reactance when considering the smoothshape of current IA. Also note the similarity regarding boththe theoretical and experimental waveforms shown inFigs. 6 and 8, respectively. Figs. 11 and 12 display ac anddc experimental waveforms for a ¼ 08 and a ¼ 308,respectively. They consider two cases: (1) without switchesS1 and S2; and (2) with switches S1 and S2.

3.2 Transient behaviour of the prototype

The study is related to both a sudden variation of load currentand a short circuit applied to switches S1 and S2.

Fig. 13 shows current IA when there is a sudden variationof current IL from 10 to 25 A. Note the efficient behaviour ofthe control circuit to respond quickly to the new operatingpoint.


Figure 8 Experimental construction of input current IA withinductive load and a ¼ 08 (scale: 25 A/div)

Figure 9 Experimental currents Is1 and Is2 with inductiveload and a ¼ 08 (scale: 30 A/div)

469


470

&

www.ietdl.org

If a failure occurs to switches S1 and S2, a short circuitapplied to these switches deactivates them, allowing anynecessary repair to take place without disconnecting thesystem. Note in Fig. 14 when in point X the short circuitis applied and in point Y it is released.

Figure 10 Currents IA, IB and IC (20 A/div) and current IL

(10 A/div) with inductive load and a ¼ 08

Figure 11 Voltage and current on ac side and dc currentconsidering a ¼ 08a Without switchesb With switchesScales: VA (50 V/div); IA(10 A/div); IL(5 A/div)


4 HVdc applications4.1 Point-to-point and back-to-backHVdc systems

The two distinct types of HVdc systems being constructedare: (1) point-to-point and (2) back-to-back [10]. Theconventional configuration used in point-to-point HVdcsystems is shown in Fig. 15. A pulse number of 12 isobtained by using two series-connected Graetz bridges,driven with ac inputs shifted by 308 through Y/Dtransformer connections. The configuration used in back-to-back HVdc systems is similar to the one shown inFig. 15, but without the dc-line.

This paper shows the possibility of using modifiedconverters with high-quality input/output waveforms andtherefore eliminating the need for filtering at both ac anddc sides. Particularly, ac filters have a complex designand performance and constitute a considerable part ofthe volume and cost of present dc terminal stations.Furthermore, in conventional HVdc rectifiers, an overlap

Figure 12 Voltage and current on ac side and dc currentconsidering a ¼ 308a Without switchesb With switchesScales: VA(50 V/div); IA(10 A/div); IL(5 A/div)


IETdo

www.ietdl.org

conduction of thyristors takes place, and depending on thetransformer leakage reactance, two, three or four valves mayconduct at any time [10]. This is particularly complicatedat the inverter end, because, due to the overlap period, theextinction angle g is considerably reduced, giving way topotential commutation failures. Conversely, the proposedconverter presents no overlap period regarding thyristorconduction. Also, it offers great flexibility and scope foroptimisation in the design process, especially regardingreactive power requirements. This area, however, must berigorously evaluated.

4.2 Directly connected HVdc generatingstations

The conventional arrangement of a hydro power station thatfeeds power into an HVdc transmission line is shown inFig. 16a. The basic feature of this scheme is the parallelingdone on the high-voltage side of the generator

Figure 13 Behaviour of current IA when different levels ofload current IL are applied (IL is shown in average value)

Scales: IA(10 A/div); IL(10 A/div)

Power Electron., 2009, Vol. 2, Iss. 4, pp. 466–473i: 10.1049/iet-pel.2007.0014

transformers. The ac harmonic filters are also connected tothis point. A simplification of the previous configuration, asproposed by some authors [11], is shown in Fig. 16b. If thegenerating and rectifying stations are intimately connected,as proposed, there are significant and obvious cost benefitswhich accrue in the elimination of many components suchas generator step-up transformers, high-voltage acswitchyard, filters and several ac collector lines between thepowerhouse and the ac switchyard. Reliability is alsoimproved because the absence of filters eliminates resonanceproblems and the risk of generator self-excitation. Moreover,there is potential for variable speed operation to optimisegeneration efficiency as a function of loading and water headconditions and potential also for an overall optimisation ofthe machine set (generator plus turbine) with respect to theelectrical frequency and mechanical speed.

There have been, however, technical concerns raised.Previous studies have shown that the scheme may notoperate satisfactorily without the ac filters [12]. If the acfilters are eliminated, harmonic currents flow into the

Figure 15 Conventional point-to-point HVdc system

Figure 14 Behaviour of IA when switches S1 and S2 areshort circuited (Scale: 10 A/div)

471


47

&

www.ietdl.org

generators with the following detrimental effects on theirperformance:

† The generator subtransient reactance forms a part of thecommutation reactance, thereby increasing the rectifiercommutation angle.

† There are both additional rotor-heating and pulsatingelectric torque.

The proposed rectifier, by not presenting input/outputharmonics and overlap angle, may give way to the modifiedinterconnection proposed by some authors [11], butreducing considerably the technical concerns raised byothers [12]. In this case, the generators have to satisfy thereactive power requirements of the rectifier; however,generators are normally rated with a low nominal powerfactor.

Figure 16 HVdc generating stations

a Conventional stationb Integrated station

2The Institution of Engineering and Technology 2009

5 ConclusionThe main advantages of the proposed rectifier are: (1) unitypower factor if diodes are used; (2) high-quality current/voltage waveforms at both sides of the rectifier; (3) simplecontrol system able to deal with both rapid load variationsand switch failures; and (4) absence of overlap conductionof thyristors. The results are sufficiently encouraging tomerit further research regarding industrial applications;especially those related to HVdc systems (point-to-point,back-to-back or directly connected generation). In thesesystems, by using the proposed rectifier, both ac and dcfilters are eliminated and also the absence of thyristoroverlap angle may be of great importance regardingcommutation failures at the inverter end. The proposedrectifier is being implemented on a typical HVdc system(CIGRE benchmark model) to study different technicalaspects, which will be reported in the near future.

6 AcknowledgmentThe authors would like to express their most sincere gratitudeto those people, listed in the references, who in the past yearshave greatly contributed to the advancement of the techniqueapplied in this paper.

They would also like to express their gratitude tothe Chilean National Commission for Scientific andTechnological Research (Fondef project no. D01I1099) fortheir financial support.

7 References

[1] BIRD B., MARSH J., MCLELLAN P.: ‘Harmonic reduction inmultiplex convertors by triple-frequency current injection’,Proc. IEE, 1969, 116, (10), pp. 1730–1734

[2] AMETANI A.: ‘Generalised method of harmonic reductionin ac/dc convertors by harmonic current injection’, Proc.IEE, 1972, 119, (7), pp. 857–864

[3] BAIRD J., ARRILLAGA J.: ‘Harmonic reduction in dc ripplereinjection’, Proc. IEE, 1980, 127, (5), pp. 294–301

Figure 17 Control circuit for self-commutated switches S1 and S2


IEdo

www.ietdl.org

[4] ARRILLAGA J., VILLABLANCA M.: ‘24-Pulse HVdc conversion’,Proc. IEE-C, 1991, 138, (1), pp. 57–64

[5] VILLABLANCA M., ARIAS M., ACEVEDO C.: ‘High-pulse seriesconverters for HVdc systems’, IEEE Trans. Power Deliv.,2001, 16, (4), pp. 766–774

[6] VILLABLANCA M.: ‘Chilean Patent Application 1151-2003,2003

[7] VILLABLANCA M.: ‘Chilean Patent Application 1335-2004,2004

[8] VILLABLANCA M.: ‘European Patent Application04076648.7, 2004

[9] VILLABLANCA M.: ‘United States Patent Application 10/860,661, 2004

[10] KIMBARK E.W.: ‘Direct current transmission-Vol. I’ (Wiley-Interscience Press, 1971, 1st edn.)

[11] INGRAM L.: ‘A practical design for an integrated HVdcunit-connected hydro-electric generating station’, IEEETrans. Power Deliv., 1988, 3, (4), pp. 1615–1621

[12] PANG C.X., TARNAWECKY M.Z.: ‘Generator winding I2 R lossesand harmonic interference under variable frequencyoperation of an HVdc unit-connection schemes’, IEEETrans. Energy Convers., 1995, 10, (1), pp. 133–139


8 AppendixFig. 17 shows a photograph of the control circuit for self-commutated switches S1 and S2, which highlights both thesize and simplicity of the circuit. In the following, the peakvalue IM of currents IS1 and IS2 is now calculated as afunction of the average load current IL.

From Fig. 2

IL ¼ IS1 ¼ IS2 (average values) (4)

From Fig. 4

IL ¼2

p=3

ð308

08Ix sin u du (5)

where

IX ¼IM

sin 308¼ 2 � IM (6)

Then

IL ¼12 � IM

p

ð308

08sin u d u ¼

12 � IM

p� 0:134 (7)

Finally

IM ¼ 1:954 � IL (8)

473


Harmonic improvement in 12-pulse series-connected line-commutated rectifiers

Documents