Av

SS

a

ARRAA

KUDHL

1

ayeivp

ewiptUsodd

aD

oT

0d

Electric Power Systems Research 79 (2009) 255–264

Contents lists available at ScienceDirect

Electric Power Systems Research

journa l homepage: www.e lsev ier .com/ locate /epsr

ctive power flow control in a distribution system using discontinuousoltage controller

achin Goyal, Arindam Ghosh ∗, Gerard Ledwichchool of Engineering Systems, Queensland University of Technology, Brisbane, Qld 4001, Australia

r t i c l e i n f o

rticle history:eceived 4 December 2007eceived in revised form 14 May 2008ccepted 6 June 2008

a b s t r a c t

This paper proposes a hybrid discontinuous control methodology for a voltage source converter (VSC),which is used in an uninterrupted power supply (UPS) application. The UPS controls the voltage at thepoint of common coupling (PCC). An LC filter is connected at the output of the VSC to bypass switching

vailable online 16 July 2008

eywords:PSiscontinuous current

harmonics. With the help of both filter inductor current and filter capacitor voltage control, the voltageacross the filter capacitor is controlled. Based on the voltage error, the control is switched between currentand voltage control modes. In this scheme, an extra diode state is used that makes the VSC output currentdiscontinuous. This diode state reduces the switching losses. The UPS controls the active power it suppliesto a three-phase, four-wire distribution system. This gives a full flexibility to the grid to buy power from

g onPSCA

ptctuipIrttct

Tccii

ybrid controlC output filter and power flow control

the UPS system dependinthrough simulation using

. Introduction

The power quality issues such as harmonics, unbalancing, volt-ge sag/swell have gained considerable attention over the last fewears. In addition, the concern for global warming has seen consid-rable development in the distributed generation (DG) technologyn the recent time. An uninterruptible power supply (UPS) can pro-ide solution for the power quality problem, while supplying cleanower from DGs at the same time.

Usually UPSs are used to provide battery back up to comput-rs and other data processing devices. They can be off-line type inhich they remain idle until power failure occurs and start supply-

ng power thereafter. The on-line type UPS continuously powers therotected load from its reserve while simultaneously replenishinghe reserve by drawing power from the utility supply. A bi-modePS uses the ac–dc rectification for battery charging [1]. It then

eamlessly picks up the load once a power failure occurs. Vari-us configurations of UPS are reported in Ref. [2]. In Ref. [3], theesign of an LC filter that is connected at the output of the UPS is

iscussed.

In this paper, the function of the UPS has been extended. It isssumed that the UPS is supplied by a distributed generator. TheG can be a PV stack or fuel cell stack [4] or battery which sup-

∗ Corresponding author at: School of Engineering Systems, Queensland Universityf Technology, 2 George Street, GPO Box 2434, Brisbane, Qld 4001, Australia.el.: +61 7 3138 2459; fax: +61 7 3138 1516.

E-mail address: a.ghosh@qut.edu.au (A. Ghosh).

pbpwvei

wo

378-7796/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.epsr.2008.06.002

its cost and load requirement at any given time. The scheme is validatedD.

© 2008 Elsevier B.V. All rights reserved.

lies the dc bus of the voltage source converter (VSC) connectinghe UPS to the ac grid. It is also possible to use PV or fuel cell toharge batteries that constitute the dc bus [5]. It is assumed thathe UPS is connected at the point of common coupling (PCC) of thetility and a major load center. The UPS is then can be controlled

n two different modes. In Mode-1, the utility supplies a constantower to the load and the remaining power comes from the UPS.

n Mode-2, the UPS supplies a constant power to the load and theemaining power comes from the utility. An interchange betweenhe modes can occur instantaneously. However, in both the modes,he UPS provides a balanced voltage at the PCC [6], thereby can-elling out any unbalance and harmonic content of the load andherefore enhancing the power quality of the distribution system.

Two different VSC control strategies are considered in this paper.he first strategy, proposed in this paper, uses both voltage andurrent control modes. In the voltage control mode, the modifiedoncept of three-level hysteretic current control (HCC) given in [7,8]s used. Generally in a three-level HCC, the output of the converters either +Vdc, 0 or −Vdc. In the zero state, a short circuit path isrovided through converter. In this paper, the zero state is modifiedy a diode state. In the diode state, all four switches of a single-hase converter are turned off. Filter inductor current is made zeroith the help of the diodes. The switching between the current and

oltage control modes is based on the error in PCC voltage. If the

rror is large, the VSC operates in the current control mode and its switched to the voltage control mode as the error reduces.

The performance of the proposed control strategy is comparedith a three-level state feedback controller [9,10] for comparison

f total harmonic distortion (THD) of PCC voltage, conduction and

2 ystem

sP

2

UsteairP

2

VVapfit

Fs

tlidoS

vcpvicari

2

56 S. Goyal et al. / Electric Power S

witching losses. Extensive simulation studies are performed usingSCAD and some of the results are presented in the paper.

. Power flow control

The single-line diagram of a distribution system containing thePS is shown in Fig. 1. For simplicity, it is assumed that the UPS is

upplied from a battery with a dc voltage of Vdc. The load is assumedo be unbalanced and nonlinear. The load is connected at the farnd of a feeder with an impedance of R + jωL, which is supplied bysource (vs). An LC filter is connected at the output of VSC, the

nductance and the capacitance of which are denoted by Lf and Cf,espectively, while the resistance Rf indicates the circuit losses. TheCC voltage is denoted by vt.

.1. The UPS model

Fig. 2 shows the structure of the UPS. It contains three H-bridgeSCs. All three VSCs are connected to common dc storage source

dc. Each VSC is connected to grid through single-phase transformernd a capacitor. The three single-phase transformers are used torovide isolation [7]. Leakage inductance of transformer Lf and thelter capacitor Cf constitute the LC filter for each phase. All threeransformers are connected in star and neutral point is connected

Fig. 1. Single-line diagram of the distribution system containing a UPS.

ig. 2. UPS structure containing three VSCs that are supplied from a common dcource.

V

a

P

Q

wrpEiog

wi

2

smT

ı

wf

esmict

fma

s Research 79 (2009) 255–264

o the neutral of the load or it may be grounded if neutral point ofoad is not available. As far as structure of single-phase converters concerned, it has four switches with anti-parallel freewheelingiode. In the later part of this paper, we shall see that output currentf the converter, i.e., filter inductor current may be discontinuous.o diodes’ rating must be same as IGBTs’ rating.

The main purpose of the UPS is to maintain a balanced PCColtage (vt) irrespective of unbalance and distortion in the loadurrents. It also controls the magnitude of the PCC voltage to are-specified value even when there is a sag or swell in the sourceoltage. To control the power flow according to the modes describedn the previous section, the angle of the PCC voltage must beontrolled. Therefore the reference for the PCC voltage containspre-specified magnitude and an angle that is based on power

equirements. The VSC then will have to track this reference voltagen order to achieve the desired performance.

.2. Reference generation

Let us define the rms source and PCC voltages as

s =∣∣Vs

∣∣∠0◦ and Vt =∣∣Vt

∣∣∠ − ı (1)

Then from Fig. 1, we get the following expression for the activend reactive power entering the PCC from the source

s =∣∣Vt

∣∣R2 + X2

[R(∣∣Vs

∣∣ cos ı −∣∣Vt

∣∣) + X∣∣Vs

∣∣ sin ı]

(2)

s =∣∣Vt

∣∣R2 + X2

[X

(∣∣Vs∣∣ cos ı −

∣∣Vt∣∣) − R

∣∣Vs∣∣ sin ı

](3)

here X = ωL. If |Vs| ≈ |Vt|, the first term inside the bracket on theight hand side of Eq. (2) is negative. However, its influence on theositive valued second term is negligible as R � X. It is clear fromq. (2) that real power flow can be controlled using ı. Because if ı isncreased or decreased, the amount of real power can be increasedr decreased accordingly. Therefore the reference PCC voltages areiven by

vrefa = Vtm sin(ωt − ı)vrefb = Vtm sin(ωt − ı − 120◦)vrefc = Vtm sin(ωt − ı + 120◦)

}(4)

here Vtm is a pre-specified magnitude and ı is controlled accord-ng to the power control mode.

.2.1. Mode-1 controlIn this mode, a fixed amount of active power is drawn from the

ource and the UPS supplies the balance amount of load require-ent. As mentioned before, this can be achieved by controlling ı.

his gives the following power control loop

= Kpses + Kis

∫es dt + Kds

des

dt(5)

here es = Psref − Ps, Psref is the reference power which is drawnrom the source.

As mentioned earlier, the UPS must regulate the PCC voltageven during changed source voltage condition. Suppose a voltageags occurs causing source voltage (|Vs|) to drop. As per require-ent, the PCC voltage |Vt| is still held constant. Hence from Eq. (2),

t is evident that ı must increase in order to maintain the power flowonstant. On the contrary, the angle ı must reduce during swell in

he source voltage.

Also, it can be seen from Eq. (3) that reactive power (Qs) isunction of ı. Furthermore, since |Vs| cos ı < |Vt|, Qs is negative. This

eans that the UPS supplies reactive power to the grid. As describedbove that at the time of sag in source voltage, ı is increased to

ystems Research 79 (2009) 255–264 257

mbp

2

aH

ı

wfct

3

acc

3

tpsvs(sb

vtedrsoseispSibmc

czbiiota

ddtdffcc

s

I

wtob

If if < 0 then S1,4 = +1 and S2,3 = 0 (8)

S. Goyal et al. / Electric Power S

aintain Ps constant. This will cause the first term of Eq. (3) toecome more negative. It means source will draw more reactiveower from the UPS during this condition.

.2.2. Mode-2 controlIn this mode, a fixed amount of power is drawn from the UPS

nd the source supplies the balance amount of load requirement.ence the ı control loop is given by

= Kpueu + Kiu

∫eu dt + Kdu

deu

dt(6)

here eu = Puref − Pu, Puref being the reference power that is requiredrom the UPS. This controller works in a similar fashion as Mode-1ontroller and adjusts the angle ı to regulate the power flow fromhe UPS.

. Control strategies

Many control methodologies are available to track the PCC volt-ge. In this section, we shall introduce a new control method andompare with an existing method. These two methods are dis-ussed below.

.1. Hybrid discontinuous control

This control scheme is useful, when an LC filter is connected tohe output of a single-phase converter and the instantaneous out-ut voltage of filter capacitor (vcf) needs to be controlled. Fig. 3hows the circuit configuration of a single-phase H-bridge con-erter with the LC filter. In this control method, the converterwitching decision is made based on both filter inductor currentif) and filter capacitor voltage. The control is switched from onetage that is based on inductor current to second stage which isased on capacitor voltage according to zones.

The zonal scheme is shown in Fig. 4 for a sinusoidal referenceoltage with zones and hysteresis band. The controller is assumedo be in Zone-1 when the error is large and it is in Zone-2, when therror is small (within the hysteresis band). In Zone-1, the controlecision is made based on filter inductor current. When the erroreduces as a result of the control action of Zone-1, the controllerwitches to Zone-2, where the decision will be made dependingn filter capacitor voltage. However, in both the zones, all threetates (+1, −1 and 0) of converter are used based upon the refer-nce voltage. In +1 state, the converter output becomes +Vdc, whichs achieved by turning on the switches S1 and S4 and turning off thewitches S2 and S3 of Fig. 3. Similarly in −1 state, the converter out-ut becomes −Vdc that can be achieved by turning on the switches

2 and S3 and turning off the switches S1 and S4. Usually the 0-states defined when output voltage of converter is 0 V which is achievedy closing the pair S1–S2 or S3–S4. But in this paper, the 0-state isodified be a diode state where all the four switches S1–S4 of the

onverter are turned off. During this time, the anti-parallel diodes

Fig. 3. VSC scheme to control capacitor voltage.

Ld

Fig. 4. The two zones voltage control scheme.

onduct to bring filter inductor current (if) to zero. Once if becomesero, the diodes stop conducting and the converter output voltageecomes equal to filter capacitor voltage (vcf). Because of the major

nvolvement of the diodes, this state is termed as the diode staten this paper. This state introduces discontinuity in the converterutput current (if). Since the converter current (if) becomes zero inhe diode state, the next switching for either +1 or −1 state, occurst zero current resulting in a loss free switching.

To track a sinusoidal capacitor voltage reference, one cycle isivided in four regions as shown in Fig. 5 where each region isivided into two zones explained above. For each zone, the hys-eresis band is defined separately. In Zone-1, the hysteresis band isefined for the inductor current and in Zone-2 the band is definedor the capacitor voltage. In Regions 1 and 3, the capacitor is chargedor positive and negative voltage, respectively. Thus the switchingontrol logic is very much similar in Regions 1 and 3. Only referencehanges from positive to negative.

If the reference of the capacitor voltage is in Region 1, thenwitching in Zone-1 will be defined as

f if ≥ hi then S1−4 = 0 (7)

here hi is the upper current hysteresis band. This implies that ifhe current if is greater than or equal to hi, turn all the switchesff. However, if this current is less than zero which is lower currentand limit, apply positive voltage at the VSC output, i.e.:

et the upper and lower voltage hysteresis bands (for Zone-2) beenoted by huv and hlv, respectively. Then for Zone-2, we have the

Fig. 5. Four regions of one cycle of a sinusoidal wave.

258 S. Goyal et al. / Electric Power Systems Research 79 (2009) 255–264

d (b) i

f

I

I

Tti

I

I

T

I

I

Fmicaai

I

I

T

I

I

S

I

I

T

I

I

Iocce

ooo

hitcaatifidco

C

TTHPtW

Vtdtititdt1rbaFtcfvoaw

Fig. 6. (a) Capacitor voltage an

ollowing switching logic:

f vcf ≥ vref + huv then S1−4 = 0 (9)

f vcf ≤ vref − hlv then S1,4 = +1 and S2,3 = 0 (10)

he switching for Region 3 can also be defined similarly. Since inhis region, the inductor current would be negative, the switchingn Zone-1 for Region 3 is defined as

f if ≤ −hi then S1−4 = 0 (11)

f if > 0 then S1,4 = +0 and S2,3 = 1 (12)

he switching in Zone-2 would be

f vcf ≤ vref − huv then S1−4 = 0 (13)

f vcf ≥ vref + hlv then S1,4 = +0 and S2,3 = 1 (14)

or Regions 2 and 4, a separate switching scheme needs to be imple-ented because the capacitor discharges in these regions. While

n Region 2, the capacitor discharges from positive to zero, it dis-harges from negative to zero in Region 4. The two zones mentionedbove are also defined for Regions 2 and 4, one for current controlnd the other for voltage control. For Region 2, switching in Zone-1s

f if ≥ 0 then S2,3 = +1 and S1,4 = 0 (15)

f if ≤ hi then S1−4 = 0 (16)

he switching for Zone-2 is

f vcf ≥ vref + hlv then S2,3 = 1 and S1,4 = 0 (17)

f vcf ≤ vref − huv then S1−4 = 0 (18)

imilarly the switching in Zone-1 for Region 4 is defined as

f if ≤ 0 then S2,3 = +0 and S1,4 = 1 (19)

f if ≥ hi then S1−4 = 0 (20)

he switching for Zone-2 is

f vcf ≥ vref + huv then S1−4 = 0 (21)

f vcf ≤ vref − hlv then S2,3 = 0 and S1,4 = 1 (22)

t is to be noted that, in Region 1, even if all four switches are turnedff, the capacitor voltage increases until if becomes zero forcing theapacitor voltage to become more than vref + huv. If huv and hlv arehosen to be the same, a steady state error will always exist. To

liminate this error, huv should be chosen less than hlv.

The key factor for satisfactory operation of the system dependsn the filter parameters Lf and Cf. For the discontinuous mode ofperation, a small value of Lf has to be chosen such that the inverterutput current has a high di/dt. Thus the small value of the inductor

tz(

a

nductor current for Example 1.

elps in charging the capacitor rapidly. However a small value ofnductor increases the filter current, which can be controlled withhe help of controller algorithm. Large inductor will lead towardsontinuous current at switching and less controlled capacitor volt-ge. Moreover, large inductor will have more energy stored in it andfter the closing the all four switch, it will transfer this energy tohe capacitor. This will result in unwanted voltage rise in the capac-tor, which is undesirable. Furthermore, it must be ensured that thelter inductance Lf and the capacitor Cf do not resonate at the fun-amental frequency. Let for the chosen value (Lf) of the inductor, aapacitor with a value of Cf0 resonates at the fundamental frequencyf ω, such that

f0 = Lf

ω2(23)

he value of the filter capacitor Cf should never be chosen near Cf0.his can be achieved by choosing either Cf � Cf0 or Cf � Cf0 [11].owever, if Cf is chosen to be large, the impedance between theCC and ground becomes very small resulting in excessive currentshrough filter capacitors. Therefore the choice of Cf � Cf0 is invalid.

e must therefore restrict Cf to be much smaller than Cf0.Example 1: In this example, a passive load is connected to the

SC through an LC filter and it is desired that capacitor voltageracks the reference voltage 1.0 sin(100�t). For a passive load, aiode state is required for all the regions shown in Fig. 5. In addi-ion to the 0-state, only +Vdc is required in Regions 1 and 4, whilen Regions 2 and 3, only −Vdc is required. Note that there is no needo impress −Vdc in Regions 1 and 4 and +Vdc in Regions 2 and 3. Butf the passive load is small and it discharges the capacitor rapidly,hen +Vdc may be required in Regions 2 and 3. Hence the schemeescribed above produces very accurate tracking. However whenhe load is active in nature, −Vdc may also be required in Regionsand 4 +Vdc may be required in Regions 2 and 3 depending on the

elative phase difference between the VSC output voltage and theack emf. This may increase the switching frequency of the VSCnd lead to continuous conduction mode of the inductor current.ig. 6 shows the load voltage and inductor current which is discon-inuous in nature for a passive load. Fig. 7 shows the filter inductorurrent with switching signals for switches S1 and S4. It is clearrom the figure that when switches S1 and S4 are on, the outputoltage of VSC is +Vdc and when these two switches are turnedff, all four switches are off because the switches S2 and S3 werelready off. Thus the inductor current reduces and becomes zeroith the help of diode conduction. During the diode conduction,

he output voltage becomes −Vdc and as inductor current becomesero, the output voltage becomes equal to filter capacitor voltagevcf).

Since every +1 or −1 state is followed by a diode state, it results inlternate switching at zero current. Hence half of the total switching

S. Goyal et al. / Electric Power Systems Research 79 (2009) 255–264 259

io

t

I

I

B

I

I

I

T−

i

I

I

F

I

I

I

S2t

3

x

w

x

u

wirpumlI0s+Smv

Fig. 7. Inductor current with diode state and VSI output voltage.

s free of losses. Fig. 8 shows that one switching edge (either risingr falling) occurs at almost zero current.

For tracking in Region 1 with an active load, the switching con-rol is defined as follows. If capacitor voltage is in Zone-1, then

f if ≥ hi then S1−4 = 0 (24)

f if < 0 then S1,4 = 1 and S2,3 = 0 (25)

ut if the capacitor voltage is in Zone-2, then

f vcf ≤ vref − hlv then S1,4 = 1 and S2,3 = 0 (26)

f vcf > vref − hlv and vcf < vref + huv then S1−4 = 0 (27)

f vcf ≥ vref + huv then S1,4 = 0 and S2,3 = 1 (28)

he switching for Region 3 will be the same as above except thatVdc state will be chosen instead of + Vdc state.

The switching for Region 2 two is defined as follows. The switch-ng in Zone-1 is

f if ≥ 0 then S2,3 = 1 and S1,4 = 0 (29)

f if ≤ hi then S1−4 = 0 (30)

or Zone-2,

f vcf ≥ vref + huv then S2,3 = 1 and S1,4 = 0 (31)

f vcf < vref + huv and vcf > vref − hlv then S1−4 = 0 (32)

f vcf ≤ vref − hlv then S2,3 = 0 and S1,4 = 1 (33)

witching for Region 4 can be obtain in the same way as in Region, the only difference being the capacitor discharges from negativeo zero instead of positive to zero.

Fig. 8. Inductor current and switching.

sT0o

4

Ptwm

t(u

wa1

Fig. 9. Three-level modulation with state feedback control.

.2. State feedback control

From Fig. 3, the state space equation of the system is given by

˙ = Ax + Buc (34)

here

=[

ifvcf

], A =

⎡⎣−Rf

Lf− 1

Lf1Cf

0

⎤⎦ , B =

[Vdc

Lf0

]

The state feedback control law is then given by

c = −K

[ifHP − 0

vcf − vref

](35)

here K is the feedback gain matrix and ifHP is obtained by passingf through a high pass filter. Since it is rather difficult to obtain aeference for if, it is safer to enforce that the high frequency com-onents of this signal should be zero. The gain matrix K is obtainedsing a linear quadratic regulator (LQR) design, where the maxi-um weightage is given to vcf. Once uc is obtained, the switching

ogic is generated using a three-level HCC control as shown in Fig. 9.n this figure, two hysteresis bands are chosen. Suppose uc starts at

and crosses the upper band h1 and reaches point A. Then thewitches S1 and S4 are turned on such that the output voltage isVdc. When uc reaches −h1 at point B, the output voltage is made 0.ubsequently, when uc reaches h1 at point C, the output voltage isade +Vdc. Thereafter when uc reaches −h1 at point D, the output

oltage is made 0. If uc continues in the same direction and over-hoots −h1 to reach point D, then the output voltage is made −Vdc.hereafter when uc reaches h1 at point F, the output voltage is made. This logic is applied depending on the current and previous statef uc.

. UPS performance evaluation

Once the reference for PCC voltage is computed from Eq. (4), theCC voltage can be tracked using the controllers described in Sec-ion 3. Application of hybrid discontinuous controller is illustratedith the help of the following example where a comparison is alsoade with the state feedback controller.Example 2: Consider the system shown in Fig. 1. It is assumed

hat the source voltage vs is balanced and has a magnitude of 11 kVL-L, rms), which is equal to about 9 kV (L-N peak). The load is annbalanced nonlinear load, whose unbalanced RL components are

Zla = 48.2 + j94.2 ˝, Zlb = 12.2 + j31.4 ˝

and Zlc = 24.2 + j60.47 ˝

here the subscripts a, b and c denote the three phases. In addition,three-phase rectifier is also connected to the PCC with an output of00 + j31.4 �. The feeder impedance is R + jωL = 3.025 + j18.13 �. The

260 S. Goyal et al. / Electric Power Systems Research 79 (2009) 255–264

g load

UpiitpcUflts(taFcrptvssa

cTHPsMcatbmsfccn

t

TC

SD

Fig. 10. UPS performance durin

PS parameter are Vdc = 3.5 kV, Rf = 1.5 � and Cf = 50 �F. The single-hase transformers are rated 1 MVA, 3 kV/11 kV with a leakage

nductance of 2.5%. The UPS is connected to grid at time t = 0 wheret is required to maintain the PCC voltage at 11 kV (L-L rms). At= 0.3 s, one more balanced three-phase load with 50 + j94.2 � (perhase) is connected. It is assumed that the source power remainsonstant (Mode-1 control) such that balance power comes from thePS. The additional load is turned off at t = 0.5 s. The active powerowing through the circuit is shown in Fig. 10(a). It can be seenhat the source real power remains constant during the load tran-ients. The PCC voltage angle is shown in Fig. 10(b). Note from Eq.4) that the PCC voltage angle is −ı. Hence in the figure and in allhe subsequent figures, the PCC voltage angle indicates −ı. Thislso remains constant all through the load changes, as expected.rom Eq. (3), it is obvious that if the voltage and angle ı remainonstant, the reactive power drawn from the source should alsoemain constant. This is shown in Fig. 10(c) where the UPS reactiveower changes in sympathy with that of the load, while main-aining the source reactive power constant. The three-phase PCC

oltage during the load increase is shown in Fig. 10(d). It can beeen that this voltage does not get disturbed during the load tran-ient. The controller parameters chosen are Kps = −0.025, Kis = −10.0nd Kds = −10 × 10−6.

a

tg

able 1omparison of state feedback and discontinuous control

Switching edges in one cycle (0.02 s) THD (PCC voltage

tate feedback 914 2.4iscontinuous control 286 0.68

transient with Mode-1 control.

To compare the results obtained with the hybrid discontinuousontrol, the same test is repeated with the state feedback control.he results obtained are very similar and hence are not shown here.owever, the switching and conduction losses and the THD of theCC voltages are computed for both these methods. To compute thewitching losses, it is assumed that the VSC is made of Toshiba GTRodule Silicon N Channel IGBT (MG400J1US51) IGBTs, the specifi-

ations for which are given in Appendix A. The comparative resultsre listed in Table 1. It can be seen that the switching edges inhe discontinuous control is only 31% of those in the state feed-ack control. The current through switches in the discontinuousode is higher than that in the state feedback control. Hence the

witching losses in the discontinuous control are 56% of the stateeedback control and the conduction losses are higher in the dis-ontinuous control. It can therefore to be concluded that the highurrent through switches in the discontinuous control somewhatullifies the benefit of less switching edges.

The state feedback controller is not used in the remaining part ofhe paper. Only the results with the hybrid discontinuous controller

re presented subsequently.

Example 3: To investigate the behavior of the UPS controller athe time of sag in source voltage, we consider the same system asiven in Example 2. With the system operating in steady state, an

) (%) Conduction losses in switches (J), 0.02 s Switching losses (J), 0.02 s

7.41 166.1410.47 93.18

S. Goyal et al. / Electric Power Systems Research 79 (2009) 255–264 261

Fig. 11. System performance at the time of asymmetrical sag in source voltage for Example 3.

Fig. 12. Mode-2 control at the time of sag and swell in the source voltage for Example 4.

Fig. 13. Mode-2 control during power reference change.

2 ystems Research 79 (2009) 255–264

uvpjNabrlseEssbwbitv

mtuccasTKiovp

trad

t1Tct

62 S. Goyal et al. / Electric Power S

nsymmetrical voltage sag is created at t = 0.3 s, where the sourceoltage of the phases a, b and c decreases to 7, 6 and 8 kV (L-N,eak), respectively. The angles of all the phase voltages also make a

ump of −10◦. The UPS however regulates the PCC voltage at 9 kV (L-, peak). Fig. 11(a) shows the three-phase source voltage with thesymmetrical sag. Fig. 11(b) shows the three-phase PCC voltagesefore and after the occurrence of the sag. It can be seen that peaksemain constant barring slight distortion at the onset of the sag thatasts less than a cycle. The active power supplied and consumed ishown in Fig. 11(c). It can be seen that power remains constantxcept for oscillatory transients that last less than 10 cycles. Fromq. (2), it is clear that if source voltage reduces, the angle differencehould increase to maintain the same level of power flow. Fig. 11(d)hows that the angle ı. It can be seen that the angle differenceefore the sag was around 23◦ (since the angle of the source voltageas 0◦). Due to the sag, the angle of the source voltage has retarded

y 10◦. Since ı settles at −45◦, the net angle difference now is 35◦,.e., an increase of 8◦ in the angle difference has occurred to offsethe drop in the source voltages. The results with a swell in the sourceoltage are very similar and are not shown here.

Example 4: Let us now consider Mode-2 control. This controlode can be beneficial when running cost of the UPS is cheaper

han the grid power. At this time maximum constant power (basedpon the UPS rating) is supplied by the UPS. In other words, wean say that the UPS is operated at full load and rest of the power orhange in load power is supplied by the source. However the unbal-nced and nonlinear portion of the load is still supplied by the UPSuch that source current can be maintained balanced sinusoidal.he controller parameters are chosen as Kpu = 0.025, Kiu = 10.0 and

du = 10 × 10−6. To investigate the behavior of Mode-2 control dur-

ng sag and swell in the source voltage, it is assumed that a sagccurs at t = 0.3 s where it reduces to 6.98 kV (L-N, peak) and aoltage swell occurs at t = 0.5 s where it increases to 10.98 kV (L-N,eak). The results are shown in Fig. 12. It shows the real powers

isso

Fig. 15. System performance during m

Fig. 14. Control switching at the time of mode shifting.

hrough the different parts of the circuit. Since the load poweremains constant, the powers supplied by the UPS and the sourcelso remain constant. However the angle ı changes to accommo-ate this as evident from Fig. 12(b).

Additionally to test the power tracking behavior of the con-roller, the reference power from the UPS is changed from 2 MW toMW at t = 0.3 s. The additional power must come from the source.his is evident from Fig. 13(a). The angle ı is shown in Fig. 13(b). Itan be seen that the angle quickly settles to a new value to increasehe power flow from the source.

Example 5: In this example, we explore the mode shifting behav-or of the total system. This can be useful for economic operation theystem in which the maximum power is drawn from the cheaperource. The mode shifting is achieved by controlling ı with the helpf two PID controllers as shown in Fig. 14. Both the controllers are

ode switching and load change.

S. Goyal et al. / Electric Power Systems Research 79 (2009) 255–264 263

e duri

rsoStapStiEtoa

iltlswdaFaU

icctpnsa

thpscdaitti

tcltvaweantt

5

sS

Fig. 16. System performanc

un simultaneously irrespective of the control mode. For example,uppose the control is running in Mode-1 with the power referencef Psref. Then ı is computed from Eq. (5) with the switches Sw1 andw3 connected in the upper position. In order to avoid any largeransient in case of a mode switch, Mode-2 controller is also runt this time. However, the reference for Mode-2 controller is com-uted from the balance power, i.e., PL − Psref. Therefore the switchw2 is also in the upper position. However since the switch Sw3 is inhe upper position, the angle ı computed from Mode-2 controllers given to the system for the computation of the voltage referenceq. (4). Once the mode shift to Mode-2 control is desired, all thehree switches are connected to the lower position. In this case theutput angle of Mode-2 with the reference of Puref is used in Eq. (4)nd Mode-1 controller runs off-line with a reference of PL − Puref.

The results are shown in Fig. 15. It is assumed that the systems operating in Mode-1 when a load increase occurs at t = 0.2 s. Theoad is reduced back to its nominal value at t = 0.3 s. Subsequentlyhe control is switched to Mode-2 at t = 0.4 s. Following this, theoad again increases at t = 0.6 s and reduces at t = 0.8 s. Fig. 15(a)hows the three active powers. It is evident that the control schemeorks in a desired manner. The angle ı is shown in Fig. 15(b). Itoes not have any significant transient barring a momentary jumpt the switching point. The source current of phase-a is shown inig. 15(c). It can be seen that it remains constant for Mode-1 controlnd changes according to the load requirement in Mode-2 when thePS supplies a constant power.

Example 6: In this example, the system performance duringslanding is investigated when the UPS is operating under Mode-2ontrol. With the system operating in the steady state, the breakeronnecting the supply side with the PCC opens inadvertently at

= 0.2 s. When there is no power from source at that time total loadower is supplied by the UPS provided that total load power doesot exceed the UPS rating. The breaker recloses at t = 0.4 s. Fig. 16(a)hows the source, load and UPS power variation during islandingnd reclosing. It can be seen that the UPS power becomes equal

sac

p

ng islanding and reclosing.

o the load power during islanding, as expected. During islanding,owever, the angle of the PCC voltage is irrelevant since the totalower is supplied by the UPS and zero power is drawn from theource. Also since the source power is zero, the output of the PIDontroller keeps on increasing. This is cause severe angle transienturing reclosing. To avoid this, the angle controller is turned offt the onset of island detection and the angle is held constant atts pre-islanded value. The angle is shown in Fig. 16(b). Phase-a ofhe source and load currents are shown in Fig. 16(c). It can be seenhat load current does not undergo any substantial transient duringslanding or reclosing.

Note that fault studies are not presented here. If a fault occurs inhe feeder supplying the load, the UPS will start supplying the faulturrent, till it is isolated. During this time the PCC voltage will col-apse and a large current will flow through the converter damaginghe power electronic components. To avoid this, either the con-erter switches to a current limiting mode or the converter switchesre blocked. Moreover, in the current limiting mode, the UPS onlyill feed the fault current till it is cleared. Therefore the best strat-

gy is to block the converter once the output current shoots beyondthreshold value with a high di/dt. The converter can be recon-

ected once the breaker trips to isolate the fault. Note that withouthe presence of a series device, a shunt compensator cannot holdhe PCC voltage during a fault in the utility feeder.

. Conclusions

This paper presents a new control methodology to balance theource currents and real power sharing between the grid and a UPS.ince the PCC voltages are balanced, the current drawn from the

ource are balanced irrespective of the nature of the load. The use ofn LC filter with the UPS removes the effect of switching frequencyomponents at PCC voltage.

A new control strategy for the control of the PCC voltage is pro-osed in this paper. In this, a diode state is used instead of the

264 S. Goyal et al. / Electric Power System

Table 2Datasheet of Toshiba Silicon N Channel IGBT, MG400J1US51

Serial no. Characteristic Rating

1 Collector emitter voltage 600 V2 Gate emitter voltage 20 V3 Forward current 400 A4 Turn on time (max) 1.2 �s5 Rise time (max) 0.3 �s6789

cctlsa

ctc

A

iE

R

[

Turn on delay (max) 0.4 �sTurn off time (max) 1.0 �sTurn off delay (max) 0.4 �sFall time (max) 0.3 �s

ommon 0-state. This results in discontinuity in the VSC outputurrent (i.e., the inductor current of the LC filter). In comparisono 3-level state feedback control, the proposed method gives muchess switching and the switching loss is almost half compared to thetate feedback control. Furthermore, the new control scheme gener-tes less THD in PCC voltage than the 3-level state feedback control.

The PCC voltage angle (ı) is controlled using a simple PIDontroller. Mode shifting between grid and UPS with the help ofwo PID controllers is discussed along with switching of these twoontrollers.

ppendix A

The characteristics of Toshiba MG400J1US51 module is listedn Table 2 [12]. These ratings are used for loss computation ofxample 2.

[

[

s Research 79 (2009) 255–264

eferences

[1] W. Bower, G. O’Sullivan, Bimode uninterruptible power supply compatibilityin renewable hybrid energy systems, IEEE Aerospace and Electronic SystemsMagazine 5 (8) (1990) 16–22.

[2] K. Hkachi, M. Sakane, T. Matsui, A. Kojima, M. Nakaoka, Cost-effective prac-tical developments of high-performance and multi-functional ups with newsystem configurations and their specific control implementations, IEEE PowerElectronics Specialist Conference 1 (1995) 480–485.

[3] F. Kamran, T.G. Habetler, A novel on-line ups with universal filtering capabilities,IEEE Transactions on Power Electronics 13 (3) (1998) 410–418.

[4] C. Wang, M.H. S Nehrir, Distributed Generation Applications of Fuel Cells, in:Proceedings of the Power Systems Conference: Advanced Metering, Protection,Control, Communication, and Distributed Resources 2006. PS ‘06, 2006, pp.244–248.

[5] Z. Jiang, R.A. Dougal, Control strategies for active power sharing in a fuel-cell-powered battery-charging station, IEEE Transactions on Industry Applications40 (3) (2004) 917–924.

[6] M.K. Mishra, A. Ghosh, A. Joshi, Operation of a DSTATCOM in volt-age control mode, IEEE Transactions on Power Delivery 18 (2003) 258–264.

[7] G.H. Bode, D.G. Holmes, Implementation of three level hysteresis current con-trol for a single phase voltage source converter, in: Proceedings of the PowerElectronics Specialists Conference, 2000, pp. 33–38.

[8] F. Zare, G. Ledwich, A hysteresis current control for single-phase multilevelvoltage source converters: PLD implementation, IEEE Transactions on PowerElectronics 17 (2002) 731–738.

[9] A. Ghosh, G. Ledwich, Load compensating DSTATCOM in weak AC Systems, IEEETransactions on Power Delivery 18 (2003) 1302–1309.

10] A. Ghosh, G. Ledwich, Power Quality Enhancement Using Custom PowerDevices, Kluwer Academic Publishers, 2002.

11] M.K. Mishra, A. Ghosh, A. Joshi, Load compensation for systems with non-stiffsource using state feedback, Electric Power Systems Research 67 (2003) 35–44.

12] Toshiba GTR Module Silicon N Channel IGBT, MG400J1US51, from the website,http://www.datasheetcatalog.com/datasheets pdf/M/G/4/0/MG400J1US51.shtml.