15 Modeling of a Voltage controlled VSI and Current Controlled VSI

782 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 3, SEPTEMBER 2006

Application of Voltage- and Current-ControlledVoltage Source Inverters for Distributed

Generation SystemsSung-Hun Ko, Seong R. Lee, Hooman Dehbonei, Member, IEEE, and Chemmangot V. Nayar, Senior Member, IEEE

Abstract—Voltage source inverters (VSI) have been widely usedin uninterruptible power supplies, unified power flow controllersor unified power quality conditioners, and distributed generationsystems (DGS). VSIs are inherently efficient, compact, and eco-nomical devices used to control power flow and provide qualitysupply. VSIs can be classified as voltage-controlled VSIs (VCVSIs)and current-controlled VSIs (CCVSIs), depending on their con-trol mechanism. In this paper, a detailed comparison of VCVSIsand CCVSIs for DGS applications is presented. This paper exam-ines the advantages and limitations of each control technique ina single-phase DGS, without incorporating additional hardwareand/or extra complex control techniques. Discussions on the con-cepts, hypotheses, and computer simulations of different VSIs inthe presence of different loads and conditions are presented. Theexperimental results confirm the validity of the analysis and simu-lations outlined. The paper provides design recommendations forthe use of VCVSIs and CCVSIs in various applications.

Index Terms—DC–AC power conversion, energy conversion,power conditioner, power electronics.

I. INTRODUCTION

DUE to improvement in technologies, electrical power canbe generated more efficiently and closer to the point

of consumption. Additionally, distributed generation systems(DGS) enable alternative energy sources (AES) to easily utilizeand supplement fossil fuels. Renewable energy sources (RES)(e.g., solar, wind, biomass, wave, hydropower, etc.) can play amajor role in the preservation of our underground resources andthe reduction of air pollutants. DGS based on RES have beenknown to be one of the most cost-effective, reliable, and durablepower systems to provide energy saving and noninterruptedpower with high power quality [1]–[5]. DGS can be classi-fied further into stand-alone and grid connected systems (seriesand parallel processing), according to the output of the voltagesource inverters (VSIs) and connection to other ac sources andloads [6]. Typical examples of other ac sources are any avail-able grid (strong, weak, or diesel grids) or other DGS sources.VSIs are inherently efficient, compact, and economical and offer

Manuscript received October 25, 2005; revised February 7, 2006. This workwas supported in part by the Australian Research Council under Grant LP0348994 and in part by the Postdoctoral Fellowship Program of the Ministry ofCommerce, Industry & Energy (MOCIE). Paper no. TEC-00360-2005.

S.-H. Ko and S. R. Lee are with the School of Electronic & InformationEngineering, Kunsan National University, Kunsan 573-701, Korea.

H. Dehbonei and C. V. Nayar are with the Department of Electrical andComputer Engineering, Curtin University of Technology, Perth 6854, Australia(e-mail: [email protected]).

Digital Object Identifier 10.1109/TEC.2006.877371

numerous functions that require a minimum number of powerconversions [7]–[11].

The parallel processing DGS controls power flow and qual-ity by controlling the power conversion between the dc busof bi-directional VSIs and the available grid [12], [13]. Thebi-directional VSIs can be further classified into VCVSIs andCCVSIs, depending on their control mechanism [14]. In DGS,VCVSIs use the amplitude and phase of an inverter out-put voltage relative to the grid voltage to control the powerflow [15]. In VCVSIs, the desired current flow is generatedby controlling the voltage across the decoupling inductor. TheCCVSI uses switching instants to generate the desired currentflow in the VSI’s inductor filter, using instantaneous currentfeedback [16].

There are advantages and limitations associated with eachcontrol mechanism. For instance, VCVSIs provide voltage sup-port to the load (the VSI operating as a voltage source), whileCCVSIs provide current support (the VSI operating as a cur-rent source). The CCVSI is faster in response compared to theVCVSI, as its power flow is controlled by the switching instant,whereas in the VCVSI, the power flow is controlled by adjustingthe voltage across the decoupling inductor. Active and reactivepower is controlled independently in the CCVSI, but are coupledin the VCVSI. Generally, the advantages of one type of VSIsare considered as a limitation of the other type. In this paper,a detailed comparison of VCVSIs and CCVSIs is investigatedfor DGS under various conditions. The experimental results ona scaled-down version (1 kVA) of DGS, confirm the validity oftheoretical and simulation studies. The design consideration andsummary of different VSI controls is presented in Section V.

II. PARALLEL PROCESSING DISTRIBUTED GENERATION

SYSTEMS

A typical configuration of the parallel processing DGS usinga VSI is shown in Fig. 1. This system consists of a VSI, whichis connected in parallel to the grid for a CCVSI and through adecoupling inductor for a VCVSI. It is generally expected thatthe VSI performs the following functions in DGS [7], [17]–[19]:

1) Load voltage stabilization (±5% voltage regulation) inboth parallel processing and stand-alone modes;

2) Uninterruptible power supply (UPS);3) Reactive power support—grid power conditioning includ-

ing power factor correction (>0.9) and harmonics miti-gation (THD<5%) (only in parallel processing mode) asper IEEE standard 1159 [17];

0885-8969/$20.00 © 2006 IEEE

KO et al.: APPLICATION OF VOLTAGE- AND CCVSIs FOR DISTRIBUTED GENERATION SYSTEMS 783

Fig. 1. Schematic diagram of a parallel processing DGS.

Fig. 2. The equivalent circuit diagram of a VCVSI-based DGS.

4) Active power support—load power conditioning includ-ing demand side management (DSM). In this mode ofoperation, a bi-directional VSI is responsible for control-ling the active power flow between the dc bus and the acgrid.

A. Voltage-Controlled VSIs in DGS

Fig. 2 shows the simplified/equivalent schematic diagram of aVCVSI. For the following analysis it is assumed that the outputlow pass filters (Lf and Cf ) of VSIs will filter out high-orderharmonics generated by pulse width modulations (PWMs). Thedecoupling inductor (Xm ) is an essential part of any VCVSI asit makes the power flow control possible. In a VCVSI, the powerflow of the DGS is controlled by adjusting the amplitude andthe phase [power angle (δ)] of the inverter output voltage withrespect to the grid voltage. Hence, it is important to considerthe proper sizing of the decoupling inductor and the maximumpower angle to provide the required power flow when designingVCVSIs.

Assuming the maximum permissible voltage fluctuation inthe grid voltage (Vg) is ±20% and the grid has to supply theactive power demanded by a resistive load, the phasor diagramof the VCVSI-based DGS is shown in Fig. 3. In this figure, it isassumed that the voltage of the inverter has to be kept constant(Vc1 = Vc2 = Vc3, load voltage stabilization). Fig. 3 shows thatas the VCVSI voltage retains constant, any changes in the gridvoltage to control the desired power flow, the power angle hasto change in proportion. The power angle could be both laggingor leading, providing either the active power flow from the gridto the VCVSI or vice versa. Fig. 3 shows that the lagging powerangles result in active power from the grid towards the inverter,regardless of the grid voltage’s amplitude and minimum powerangle obtains when the grid and the VCVSI voltage are iden-tical. This figure shows that reactive power always flows fromthe higher voltage source to the lower voltage source. Hence,the higher voltage source has to supply all the reactive powerdemanded by the decoupling inductor as well as load. In weak

Fig. 3. Phasor diagram of a VCVSI-based DGS with resistive load and gridis responsible for supplying the active power.

grid applications, when the grid voltage drops considerably, theVCVSI has to supply both the rated active power and full re-active power, resulting in over sizing of the inverter (>100%of the rated power). Unity power factor operation (Igp1 = Ig1)is only possible if the grid voltage is reduced to Vg1 and at thespecific power flow corresponds to Vx1. This is a special case,which depends on the size of the decoupling inductor, the loadand maximum permissible power angle. Therefore, power fac-tor correction is not possible using VCVSIs in DGS. This is oneof the main drawbacks of VCVSI-based DGS.

Using Fig. 2 the fundamental grid current can be expressedas (1)

Ig =Vg 0 − Vc δ

jXm= −Vc sin δ

Xm− j

Vg − Vc cos δ

Xm

(1)

where Vg and Vc are, respectively, the grid and theVCVSI’s fundamental voltages, and Xm is the decou-pling inductor impedance. Using per unit values (Sbase =V 2

base/Zbase, Vbase = Vc and Zbase = Xm) where Vbase, Zbase,and Sbase are the base voltage, impedance, and complex powervalues, respectively, the grid apparent power can be expressedas (2):

Sgpu = −Vgpu sin δ + j(V 2

gpu − Vgpu cos δ). (2)

Using per unit values, the complex power of the VCVSI anddecoupling inductor are

Scpu = −Vgpu sin δ + j[Vgpu cos δ − 1] (3)

Sxpu = j(V 2

gpu − 2Vgpu cos δ + 1)

(4)

where Sgpu, Scpu, and Sxpu are per unit values of the grid,VCVSI and decoupling inductor apparent power respectively,and Vgpu is the per unit value of the grid voltage.

As addressed above, since the load voltage must remain con-stant (load voltage stabilization), the only controllable parameterin the VCVSI is the power angle (δ). Hence, the power angle isused in a VCVSI for DSM operation. For DSM operation, it isimportant to extract the maximum power from RES and supplythis power to the load or DGS. Assuming that both RES and thegrid are supplying the demanded active power by the load, the


Fig. 4. Control block diagram of the VCVSI-based DGS used in the simula-tions and experiments.

Fig. 5. The equivalent circuit of a CCVSI-based DGS.

power angle can be calculated from (5)

Pg = PL − PRES =VgVc

Xmsin δ. (5)

Thus, the power angle (δ) is

δ = sin−1

[(PL − PRES)Xm

VgVc

](6)

where PL and PRES are the load and RES active power, respec-tively. Equation (6) explains that as the available RES energy isincreased, the power angle is reduced. This means that the RES’penetration will increase. Fig. 4 shows the block diagram of aVCVSI control system based DGS used in both the simulationsand experiments. This control block diagram includes the DSMfunction (6). If the RES available energy is more than the loadconsumption, then the power angle can be leading to export thisextra active power to the DGS.

In Fig. 4,the phase locked loop (PLL) is responsible for syn-chronizing the inverter output voltage with the grid voltage. Thesampling from the load current, RES voltage and current is alsoused to generate the required power angle (δ∗ref) (for DSM op-eration). After comparing the required/reference values and theactual variables, an error signal is generated to feed a PI con-troller. After generating the desired reference signal, it is givento the PWM generator block to generate the required switchingsignals.

B. Current-Controlled VSIs in DGS

Fig. 5 shows the equivalent schematic diagram of a CCVSI.As CCVSI controls the current flow using the VSI switchinginstants, it can be modeled as a current source and there is noneed for a decoupling inductor (Fig. 5). As the current gener-

Fig. 6. Phasor diagram of a CCVSI-based DGS with inductive load.

ated from the CCVSI can be controlled independently from theac voltage, the active and reactive power controls are decou-pled. Hence, unity power factor operation is possible for thewhole range of the load. This is one of the main advantagesof CCVSIs.

As the CCVSI connects in parallel to the DGS, it followsthe grid voltage. Fig. 6 shows the phasor diagram of a CCVSI-based DGS in the presence of an inductive load (considering thesame assumption as VCVSI section). Fig. 6 shows that whenthe grid voltage increases, the load’s active power consumption,which supplied by the grid increases and the CCVSI compen-sates the increase in the load reactive power demand. In thiscase, the CCVSI maintains grid supply at unity power factor,keeping the current phase delay with respect to the grid voltageat a fixed value (Θ). Therefore, the CCVSI cannot maintain theload voltage in the presence of a DGS without utilizing extrahardware and control mechanisms. This limitation on load volt-age stabilization is one of the main drawbacks of CCVSI-basedDGS.

Assuming the load active current demand is supplied by thegrid (reactive power support function), the required grid currentcan be rewritten as follows:

I∗g = Re[IL] = Re[SL

Vg

](7)

where SLis the demanded load apparent power. For grid powerconditioning, it is preferred that the load operate at unity powerfactor. Therefore, the CCVSI must provide the remainder of therequired current (8)

Ic = IL − I∗g. (8)

In DSM, it is desirable to supply the active power by the RES,where excess energy from the RES is injected into the DGS. Theremaining load reactive power will be supplied by the CCVSI.Hence, (8) can be rewritten as (9)

I∗g = Re[IL] − Re[Ic] = Re[SL − PRES

Vg

](9)

Equations (8) and (9) show that in the worst case, the CCVSIhas to supply both the active and reactive power demanded bythe load. This means that the CCVSI sizing can be rated atfull load without the need to oversize. This is an advantage ofCCVSI-based DGS compared to the VCVSI. The control blockdiagram of the CCVSI-based DGS used in the simulation andexperiment using (9) and (10) is shown in Fig. 7.

The CCVSI control scheme samples the DGS voltage (Vg)for synchronization using a PLL. The samples of load current(ILoad), RES current (IRES), and voltage (VRES) are used to


Fig. 7. Control block diagram of the CCVSI-based DGS used in the simula-tions and experiments.

TABLE ISIMULATION CONDITIONS AND SELECTED PARAMETERS WHERE Lm IS THE

DECOUPLING INDUCTOR, Lf IS THE FILTER INDUCTOR, AND Cf IS THE FILTER

CAPACITOR

generate the desired inverter current amplitude (I∗d) (referencecurrent) using (9) and (10). After generating the reference cur-rent signal (Ic-ref)), this current is compared to the instan-taneous CCVSI current in order to generate the error current(Ierr). This error current is then given to the current regulatorblock to generate the desired instantaneous switching PWMs.

III. SIMULATION RESULTS

To compare the performance of the parallel processing DGSusing a VCVSI and CCVSI, a 1kVA system including linearand nonlinear loads was simulated using PSim software. Table Ishows the different parameters and selected values identicalwith the experimental hardware used in simulation, to providea foundation to compare results.

A. Power Conditioning in DGS

This simulation was conducted to evaluate the performanceof the different VSIs in the presence of different loads, whereVg and Vc are the voltage waveforms of the grid and inverter,and Ig, Ic, and Iload are current waveforms of the grid, inverterand load, respectively. Fig. 8 shows the power conditioning ofa DGS in the presence of an inductive load (Z = 40 36.7[Ω]),using different VSIs. Fig. 8(a) shows that the grid can supplythe load’s active power when the grid voltage is almost thesame as the VCVSI voltage at a low power angle. In this case,the required reactive power (600 var) supplied by the VCVSI,the power factor is good as the inverter size and hence thedecoupling inductor is relatively small. Fig. 8(b) shows that theCCVSI can supply the reactive power required by the load while

Fig. 8. Waveform results of power conditioning of a DGS in the presence ofan inductive load (z = 40 36.7[Ω]). (a) VCVSI. (b) CCVSI.

the active power is supplied fully by the grid in the same wayas for VCVSI.

Fig. 9(a) shows the VCVSI as a power conditioner for aDGS in the presence of a nonlinear load (a RLC diode bridgerectifier). The VCVSI cannot maintain pure sinusoidal voltageacross the nonlinear load (Vc). Hence, a portion of low-ordercurrent harmonics will be injected into the grid (Fig. 10(b),ITHD = 10.9%). Fig. 9(b) shows that the CCVSI can provideall the reactive power demanded by the nonlinear load and hencethe grid supplies only the remaining active power (unity powerfactor operation). In this case, the CCVSI prevents any low-orderharmonics from being injected into the grid (active filtering)(Fig. 10(c), ITHD = 1.1%).

Fig. 10(a) shows the DGS in the absence of VSIs. In thiscase, all the reactive power associated with low-order harmonicsfrom the nonlinear load must be supplied by the grid (ITHD =60.8%). This figure also signifies that a VCVSI cannot meet theIEEE standard 1159 (less than 5% of THD) when a nonlinearload is presented, while a CCVSI can achieve unity PF andsatisfies THD requirements of voltage and current for the fullrange of the load, without the need for an additional controller(assuming that the grid voltage is sinusoidal).

B. UPS Function in DGS

To study the performance of each VSI controller in UPSmode, the system was simulated in the presence of nonlinear


Fig. 9. Waveform results of power conditioning of a DGS in the presence ofa nonlinear load (a RLC diode bridge rectifier). (a) VCVSI. (b) CCVSI.

loads (Fig. 11). It was assumed that at 30 ms the grid fails andboth VSIs had to supply the load. As it is shown, before gridfailure the VCVSI supplied the reactive power demanded bythe nonlinear load and the rest of the reactive power demandedby the decoupling inductor supplied by the grid (when both thegrid and the VCVSI voltages were identical). This figure showsthat the VCVSI picked up the load rapidly after the grid failed[Fig. 11(a)]. Fig. 11(b) shows that the CCVSI provides/absorbsalmost all the nonlinear current to/from the load before gridfailure and full load current afterwards to supply the load. It isshown in the presence of a nonlinear load that a CCVSI cannotprovide and maintain a sinusoidal voltage waveform even withextra voltage feedback [Fig. 11(b)]. However, the VCVSI canprovide the required voltage support and UPS function, even inthe presence of a nonlinear load and without the need for extrafeedback or load estimation control algorithms.

It is shown that in the presence of a nonlinear load, the VCVSIcan maintain the load voltage VTHD at 11.2% while the loadcurrent ITHD is 50.7%. These values can be read as high as36.3% load voltage VTHD and as low as 0.7% load currentITHD in CCVSI and stand-alone operations. The significanceof this data is that both the VCVSI and the CCVSI cannotcompensate low-order harmonics of the load voltage in orderto meet IEEE standards (eg., 1159 and 944) in the presence of

Fig. 10. The harmonic spectrum of the grid current in the absence and presenceof VSIs supplying nonlinear load. (a) Without power conditioning. (b) VCVSI.(c) CCVSI.

nonlinear loads, without extra feedback and complex controlalgorithms [8].

C. DSM Function in DGS

In RES-based DGS, it is required to give priority of supplyto the RES and reduce the share of the grid to supply the load.If the RES is not enough, then both the RES and the grid willsupply the load demand. Fig. 12 illustrates the case that the gridis the only available source to supply the required active powerdemanded by the linear load (100% of 1 kw) and suddenly,when the RES becomes available (50%), allows the RES totake part and supply 50% of the load demand. Fig. 12(a) showsthat the VCVSI maintains load voltage while the RES begins tosupply 50% of the load demand. In this case, the grid current isreduced to 50% while the load current is maintained at 100%.Hence, the load does not detect any abnormality in the supply.Fig. 12(b) shows that a CCVSI can perform DSM and supplythe load (at 50%) in the same way as a VCVSI. The delay in thepower waveforms in Fig. 12 are due to the existence of a lowpass filter in the power meter used for power measurements in


Fig. 11. UPS mode waveform results of a DGS in the presence of nonlinearload. (a) VCVSI. (b) CCVSI.

PSim software. Both power diagrams in Fig. 12 show the DSMcapability of CCVSIs and VCVSIs.

D. Voltage Regulation in DGS

In this simulation, the grid voltage was changed from itsnominal value (here 200 V) to 160 V. It was assumed that thegrid has to supply the load active power. Fig. 13(a) shows thevoltage stabilization for the load when the grid voltage fluctuatedin the presence of a VCVSI. After a step down in the DGSvoltage from 200 to 160 V, the grid can still supply the activepower while the VCVSI maintains the load voltage. In this case,the VCVSI has to supply the extra reactive power demandedby the decoupling inductor, which is dependent on the DGSvoltage and the inductor’s size. In this case, the grid currentincreases due to an increase in reactive current flow from theVCVSI to the grid. Fig. 13(b) shows the voltage stabilizationfor the load when the grid voltage fluctuated in the presenceof a CCVSI. Due to the direct connection of the CCVSI to theDGS, it cannot compensate the grid voltage fluctuations withoutadditional hardware and control feedback algorithm [20], [21].Hence, the load voltage cannot be maintained and the load willsuffer from grid voltage fluctuations. As it is assumed in this casethat the DGS must supply the active power demanded by load,the CCVSI current remains at zero even after the DGS voltage

Fig. 12. Waveform results of demand side management in DGS (grid supplieslinear load from 100% to 50%). (a) VCVSI. (b) CCVSI.

drops. In this case, the load demand was reduced in proportionto the decreases in the DGS voltage, and hence resulted in lessactive power support by the DGS.

IV. EXPERIMENTAL RESULTS

Fig. 14 shows a photograph of a scale down version of aDGS that was prototyped to examine the analytical and simula-tion analysis. The experimental setup consists of a computer tomonitor and program the desired control techniques (Figs. 4 and7) into a digital signal processor (DSP), to provide a switchingsignal to a control board and a VSI. A variac is used to simulatea weak grid, while different loads are connected to the output ofthe VSIs. The scope and power analyzers were used to record theinformation for further evaluative comparisons of the analyticaland simulation results.

System specifications are given in Table II. As a battery bankof 180 V was used in this test, a low frequency transformer wasutilized to step-up the output voltage of the H-bridge inverter tothe required value (200 V).

The Voltech (PM3000A) power meter measures a power fac-tor of over 0.99 for a CCVSI and for a VCVSI, from 0.97 to0.99, depending on the load and grid voltage fluctuation forlinear loads. This relatively good power factor for the VCVSI


Fig. 13. Waveform results of voltage stabilization when the grid voltagechanges from 200 to 160 V in DGS in the presence of a linear load (R = 40 [Ω]).(a) VCVSI. (b) CCVSI.

Fig. 14. Photograph of the prototyped DGS.

TABLE IITHE SPECIFICATIONS OF THE PROTOTYPED DGS

Fig. 15. Experimental waveform results of power conditioning of a DGS inthe presence of an inductive load (z = 40 36.7 [Ω]). (a) VCVSI. (b) CCVSI.

is due to the low power operation of the DGS (1 kVA). TheTektronix (TDS3054B) digital scope was used to capture thefollowing results.

A. Power Conditioning in DGS

Fig. 15 shows that, in the presence of an inductive load(Z = 40 36.7 [Ω]), when the grid voltage is almost the sameas the VCVSI voltage, a slight adjustment to the power anglewill enable the grid to supply the load active power, and the re-quired reactive power (600 var) can be supplied by the VCVSI.Fig. 15(b) shows that at unity power factor, the grid suppliesthe load active power and the CCVSI supplies the load reac-tive power (reactive power compensation). These results alsocomply with the simulation results (Fig. 8).

Fig. 16 shows the experimental waveform results of the powerconditioning of a DGS in the presence of a diode bridge rectifierwith RLC (nonlinear load). It confirms that the CCVSI hasbetter performance in the presence of a nonlinear load forlow-order harmonic mitigation and provides unity power fac-tor operation to the DGS [Fig. 16(b)], compared to the VCVSI[Fig. 16(a)]. As was expected, the output voltage of the VCVSI(Vc) will be distorted in the presence of a nonlinear load[Fig. 16(a)]. This deformation in the voltage waveform can be


Fig. 16. Experimental waveform results of power conditioning of a DGS inthe presence of a nonlinear load (a diode bridge rectifier with RLC). (a) VCVSI.(b) CCVSI.

corrected by adding a wave-shaping control algorithm and usingextra feedback signals [8].

B. UPS Function in DGS

Experiment results for the study of the UPS function of differ-ent VSIs are shown in Fig. 17. It observed that both the VCVSIand CCVSI can supply and maintain the pure sinusoidal voltagewaveform when the grid fails, in the presence of linear loads.However, in the presence of nonlinear loads, the CCVSI can-not provide the proper voltage without additional hardware andcontrol feedback algorithms [Fig. 17(b)]. Although the VCVSIcannot mitigate low-order harmonics from the nonlinear loads,it can maintain the load voltage after the grid fails [Fig. 17(a)].These results verify the simulation results (Fig. 11). Fig. 17illustrates that after grid failure, the CCVSI can provide sinu-soidal current waveform to the load by adding predictive controlalgorithms, however, this voltage will be distorted as it tries tokeep the load current sinusoidal.

C. DSM Function in DGS

Demand side management is an important function in anyDGS which defines a load sharing among the suppliers. In thisexperiment, to study the DSM function of a VCVSI and CCVSI,it was assumed that these VSIs would suddenly have to take 50%of the load active power supply. Fig. 18(a) shows that in a DGS

Fig. 17. Experimental waveform results of UPS function of a DGS usingdifferent VSI in the presence of nonlinear load (a RLC diode bridge rectifier).(a) VCVSI. (b) CCVSI.

system using a VCVSI, where the grid and the VCVSI voltagesare identical, grid supplies the reactive power demanded by de-coupling the inductor, while the grid supplies the active power.For demand side management, the proportions of active powerto be supplied by the VCVSI and grid respectively can be con-trolled by changing the power angle (δ). In this experiment, thepower angle was modified in order that 50% of the active powerto be supplied by the VCVSI. In this situation, after the transientit was observed that without changes in the load current the gridcurrent decreased while the VCVSI current increased. Fig. 18(b)shows a DGS in the presence of a CCVSI. In this case, as thegrid was subject to supply the full resistive load there was noreactive power compensation (no decoupling inductor), hencethe converter current was maintained at zero. After a commandfrom the control system to overtake 50% of the active power bythe CCVSI, it was observed that with a very smooth transient (nochange in the load current) the CCVSI supplied the remaining50% of the load-demanded active power. These results supportthe simulation results (Fig. 12).

D. Voltage Regulation in DGS

Voltage regulation is another important feature requiredin most applications dealing with sensitive loads. Moreover,


Fig. 18. Experimental waveform results of demand side management of aDGS in the presence of a resistive load. (a) VCVSI. (b) CCVSI.

voltage stabilization would be one of the most important re-quirements in weak grid applications. The following tests werecarried out in order to study the performance of the VCVSI andCCVSI in stabilizing the voltage. In this experiment, the volt-age of the grid was changed from 200 V to 160 V (Fig. 19).Fig. 19(a) illustrates that the VCVSI shares in supplying thereactive power demanded by the decoupling inductor with thegrid, when the grid voltage is the same as the VCVSI voltage.This can be done by adjusting the power angle to allow the gridto supply all the active power demanded by the load. This ex-periment confirms the findings from the simulation in Fig. 13,namely that the VCVSI can maintain the load voltage regard-less of changes in the grid voltage. However, due to the parallelconnection of the CCVSI to the DGS, the CCVSI follows thegrid voltage and hence cannot provide voltage support to theload without extra hardware and complex control techniques[Fig. 19(b)].

V. DESIGN CONSIDERATION AND COMPARISON OF VCVSIS

AND CCVSIS IN DISTRIBUTED GENERATION SYSTEMS

The comparison of the VCVSI and CCVSI-based DGS isshown below.

Fig. 19. Experimental waveform results of voltage stabilization in a DGS inthe presence of a resistive load. (a) VCVSI. (b) CCVSI.

A. Load Voltage Stabilization

It is shown that the VCVSI can regulate the load voltagewithin ±5% as per IEEE standards (1159 and 944). In contrast,as a CCVSI is connected directly to the grid it cannot com-pensate the grid voltage fluctuation. A decoupling inductor isessential to decouple the effect of grid voltage fluctuation, whichcan be achieved by using VCVSIs. Therefore it is suggested thata VCVSI be used to provide the required voltage support to theload in applications with a weak grid [22], [23].

B. Uninterruptible Power Supply

As a VCVSI by nature performs the same as a voltage source,it can maintain voltage support for the load in the absence of agrid (stand-alone operation). It is shown that the VCVSI can-not provide a pure sinusoidal waveform in the presence of anonlinear load without extra control mechanisms and feedback.However, as is shown [8], wave shaping of the VCVSI is possi-ble with extra feedback and hence the sinusoidal output voltageis guaranteed even in the presence of nonlinear loads and instand-alone operations. On the other hand, the CCVSI cannotprovide proper voltage support as by nature it is a current sourceand voltage follower. Therefore, a VCVSI is recommended forthose applications where a UPS function is of high priority [24].


C. Reactive Power Support, Active Filtering and Power FactorCorrection

As the active and reactive powers are coupled in a VCVSI,it generally offers poor power factor correction performance atlow load, or when the grid voltage is different from the volt-age of the load/VCVSI. In contrast, a CCVSI provides goodreactive power support and decoupling from the active power.This capability enables CCVSIs to perform at unity power fac-tor and to mitigate low-order harmonics effectively. Therefore,CCVSIs are recommended for those applications where reactivepower support, including unity power factor operation or activefiltering is the main goal, (i.e., active power line conditioners(APLC) [25], [26]).

D. Active Power Support or DSM

Both the VCVSI and CCVSI offer effective bidirectionalpower flow between their dc and ac bus. The power flow controlin a VCVSI is very sensitive, and depends not only on a limitedpower angle, but also on the size of the decoupling inductor.However, as the phase and amplitude are controlled separatelyin a CCVSI, the power flow in a CCVSI is smoother due to thedecoupling of the active and reactive power in this control tech-nique. Hence, for DSM operation when voltage support is not apriority, CCVSIs are recommended (i.e., photovoltaic grid-tiedinverters).

E. Sizing the PCU

Due to the existence of a decoupling inductor, a VCVSImust supply both the active and reactive power demanded fromthe load as well as the reactive power required by the decou-pling inductor. This means that the VCVSI has to be oversized(>100%). This could be worsened by an increase in the voltageof the grid and VCVSI. However, the CCVSI can be rated atfull load (100%), as there is no decoupling inductor and it onlyneeds to supply the active and reactive power required by theload.

In practice, it is possible to change the control algorithm inVSIs with respect to the different functions required. However,there are some prerequisites as well as pros and cons associatedwith changing the control algorithm in VSIs which must beconsidered.

VI. CONCLUSION

Voltage source inverters have been widely used in many ap-plications, including distributed generation systems. VSIs areinherently efficient, compact and economical devices, whichare used to control power flow and the quality of power supply.In this paper, a detailed comparison of VCVSIs and CCVSIsfor DGS applications was presented. It was shown that neitherVCVSIs nor CCVSIs alone can offer all the functions required ina DGS. Hence, the most appropriate VSI should be chosen basedon its application and priority. Alternatively, both types of VSIsmust be used for those applications where voltage stabilization,unity power factor operation and active filtering are required.This paper examined the advantages and disadvantages of each

VSI’s control techniques in the presence of different loads andprovides design recommendations for the use of VCVSIs andCCVSIs in various applications. The experimental results verifythe theoretical analysis and computer simulations.

ACKNOWLEDGMENT

The authors are grateful to Curtin University of Technologyfor providing opportunities to carry out this research. This workwas supported in part by the Australian Research Council un-der Grant LP 0348994 and partly by the Post-doctoral Fellow-ship Program of the Ministry of Commerce, Industry & Energy(MOCIE).

REFERENCES

[1] W. Wongsaichua, W.-J. Lee, S. Oraintara, C. Kwan, and F. Zhang, “Inte-grated high-speed intelligent utility tie unit for disbursed/renewable gen-eration facilities,” IEEE Trans. Ind. Appl., vol. 41, no. 2, pp. 507–513,Mar.–Apr. 2005.

[2] T. Jin, L. Li, and K. Smedley, “A universal vector controller for three-phase PFC, APF, STATCOM, and grid-connected inverter,” presented atthe 19th Ann. IEEE Applied Power Electronics Conf. Expo. (APEC’04),2004, vol. 1, pp. 594–600.

[3] K. Dai, P. Liu, J. Xiong, and J. Chen, “Study on dual-DSP-controlled three-phase series-parallel compensated line-interactive UPS system (delta-conversion UPS),” presented at the IEEE Int. Electric Machines and DrivesConf. (IEMDC’03), Jun. 1–4, vol. 1, pp. 436–442.

[4] F. Blaabjerg, Z. Chen, and S. B. Kjaer, “Power electronics as efficientinterface in dispersed power generation systems,” IEEE Trans. PowerElectron., vol. 19, no. 5, pp. 1184–1194, Sep. 2004.

[5] M. Dai, M. N. Marwali, J.-W. Jung, and A. Keyhani, “Power flow control ofa single distributed generation unit with nonlinear local load,” presentedat the IEEE PES, Power Systems Conf. and Expo., Oct. 10–13, vol. 1,pp. 398–403.

[6] M. Prodanovic and T. C. Green, “Control and filter design of three-phaseinverters for high power quality grid connection,” IEEE Trans. PowerElectron., vol. 18, no. 1, pt. 2, pp. 373–380, Jan. 2003.

[7] A. J. Baronian and S. B. Dewan, “An adaptive digital control of currentsource inverter suitable for parallel processing inverter systems,” presentedat the IEEE 30th Ann. Meeting, Industry Applications Conf., Oct. 8–12,vol. 3, pp. 2670–2677.

[8] H. Dehbonei, “Power conditioning for distributed renewable energy gen-eration” Ph.D. Dissertation, Dept. Elect. Comput. Eng., Curtin Univ. Tech-nol., Perth, 2003.

[9] T. Kawabata, K. Ogasawara, N. Sashida, Y. Yamamoto, and Y. Yamasaki,“Parallel processing inverter system,” IEEE Trans. Power Electron., vol. 6,no. 3, pp. 442–450, Jul. 1991.

[10] H. Dehbonei, C. V. Nayar, and L. Borle, “A combined voltage controlledand current controlled ‘dual converter’ for a weak grid connected photo-voltaic system with battery energy storage,” presented at the IEEE 33rdAnnu. Power Electronics Specialists Conf. (PESC’02), Cairns, Jun. 23–27,2002, vol. 3, pp. 1495–1500.

[11] H. Dehbonei, C. V. Nayar, and L. Chang, “A new modular hybrid powersystem,” presented at IEEE Int. Symp. Industrial Electronics (ISIE’03),Rio de Janeiro, Brazil, Jun. 9–11, 2003, vol. 2, pp. 985–990.

[12] M. N. Marwali and A. Keyhani, “Control of distributed generationsystems-Part I: Voltages and currents control,” IEEE Trans. Power Elec-tron., vol. 19, no. 6, pp. 1541–1550, Nov. 2004.

[13] G. Ledwich and A. Ghosh, “A flexible DSTATCOM operating in voltageor current control mode,” IEE Proc. Gen., Transm. Distrib., vol. 149,no. 2, pp. 215–224, Mar. 2002.

[14] Z. Chen and E. Spooner, “Voltage source inverters for high-power,variable-voltage DC power sources,” Inst. Elect. Eng—Proc. Gen.,Transm. Distrib., vol. 148, no. 5, pp. 439–447, Sep. 2001.

[15] C. V. Nayar, M. Ashari, and W. W. L. Keerthipala, “A single phase un-interruptible power supply system using a bi-directional sinusoidal PWMinverter,” presented at the 1998 Int. Conf. Power Electronic Drives andEnergy Systems for Industrial Growth (PESDES’98), Dec. 1–3, 1998,vol. 2, pp. 671–676.


[16] L. J. Borle, M. S. Dymond, and C. V. Nayar, “Development and testingof a 20-KW grid interactive photovoltaic power conditioning system inWestern Australia,” IEEE Trans. Ind. Appl., vol. 33, no. 2, pp. 502–508,Mar.–Apr. 1997.

[17] Monitoring Electric Power Quality, IEEE Standard 1159, 1995. IEEEStandards Board.

[18] The Application and Testing of Uninterruptible Power Supplies for PowerGenerating Stations, ANSI/IEEE Standard 944, 1985.

[19] M. Ashari, W. W. L. Keerthipala, and C. V. Nayar, “A single phase par-allel connected uninterruptible power supply/demand side managementsystem,” IEEE Trans. Energy Convers., vol. 15, no. 1, pp. 97–102, Mar.2000.

[20] P. C. Loh, M. J. Newman, D. N. Zmood, and D. G. Holmes, “Improvedtransient and steady state voltage regulation for single and three phaseuninterruptible power supplies,” presented at the IEEE 32nd Annu. PowerElectronics Specialists Conf. (PESC’01), Jun. 17–21, vol. 2, pp. 498–503.

[21] D. N. Zmood and D. G. Holmes, “Improved voltage regulation for current-source inverters,” IEEE Trans. Ind. Appl., vol. 37, no. 4, pp. 1028–1036,Jul.–Aug. 2001.

[22] H. Dehbonei, C. V. Nayar, and L. Borle, “A multifunctional power pro-cessing unit for an off-grid PV diesel hybrid power system,” presented atthe IEEE 35th Ann. Power Electronics Specialists Conf. (PESC’04), Jun.21–25, vol. 3, pp. 1969–1975.

[23] R. L. Vasquez-Arnez and L. C. Zanetta, Jr., “Compensation strategy ofautotransformers and parallel lines performance, assisted by the UPFC,”IEEE Trans. Power Del., vol. 20, no. 2, pt. 2, pp. 1550–1557, Apr. 2005.

[24] H.-L. Jou, J.-C. Wu, C. Tsai, K.-D. Wu, and M.-S. Huang, “Novel line-interactive uninterruptible power supply,” δ/b Proc. IEE Elect. PowerAppl., vol. 151, no. 3, pp. 359–364, May 2004.

[25] P. Salmeron and J. R. Vazquez, “Practical design of a three-phase activepower-line conditioner controlled by artificial neural networks,” IEEETrans. Power Del., vol. 20, no. 2, pt. 1, pp. 1037–1044, Apr. 2005.

[26] H. C. Lin, “Intelligent neural-network-based adaptive power-line condi-tioner for real-time harmonics filtering,” IEE Proc. Gen., Transm. Distrib.,vol. 151, no. 5, pp. 561–567, Sep. 13, 2004.

Sung-Hun Ko received the B.Sc. and M.Sc. degreesfrom the Department of Control and InstrumentationEngineering, Kunsan National University, Kunsan,Korea in 1998 and 2000, respectively, and he iscurrently pursuing the Ph.D. degree in the Schoolof Electronics and Information Engineering, KunsanNational University, Kunsan, Korea.

From 2000 to 2001, he was with Research Labora-tory, Seo-Young Electronics, Inc., Korea. Currently,he is working as a Visiting Research Fellow with theDepartment of Electrical and Computer Engineering

at Curtin University of Technology, Perth, Australia. His current research inter-ests include renewable energy based distributed generation system, power factorcorrection, inverter control, and neural network.

Seong R. Lee received the B.Sc. and M.Sc. degreesin electrical engineering from Myong-Ji University,Seoul, Korea in 1980 and 1982, respectively, andthe Ph.D. degree from Chonbuk National University,Jeonju, Korea, in 1988.

From 1997 to 1998, he was the Visiting Professorwith the Department of Electrical and Computer En-gineering at Virginia Tech, VA. From 2002 to 2004,he was the Director of Engineering Research Instituteat Kunsan National University, Kunsan, Korea. Since1990, he is a Professor with the School of Electron-

ics and Information Engineering at Kunsan National University Currently, he isworking as a Visiting Professor with the Department of Electrical and ComputerEngineering at Curtin University of Technology, Perth, Australia. His current re-search interests include soft-switching inverter, power factor correction, switchmode power supply, and renewable energy based distributed generation systems.

Hooman Dehbonei (S’01–M’03) received the B.Sc.and M.Sc. degrees in electrical engineering from theIran University of Science and Technology, Tehran,Iran in 1992 and 1997, respectively, and Ph.D. de-gree from Curtin University of Technology, Perth,Australia, in 2003.

Presently, he is an Australian Research CouncilPostdoctoral Fellow with the Department of Electri-cal and Computer Engineering at Curtin Universityof Technology. He is a Chartered Professional En-gineer and National Professional Engineers Register

with the Institute of Engineers, Australia. His current research interests includepower systems (design, analysis, quality, and control), power electronics (itsapplication in power systems and renewable energy), renewable energy, andhybrid/distributed generation systems.

Chemmangot V. Nayar (M’86–SM’90) received theB.Tech. degree in electrical engineering from the Uni-versity of Kerala, India, in 1969, the M.Tech. degreein electronics from the Indian Institute of Technol-ogy, Kanpur, in 1976, and the Ph.D. degree in electri-cal engineering, specializing in wind electrical powergeneration, from the University of Western Australia,Perth, Australia, in 1985.

He holds a Personal Chair in electrical engineer-ing at Curtin University of Technology.

Prof. Nayar is a Chartered Engineer and CorporateMember of IEE, and a Chartered Professional Engineer and Fellow of IEAust.

Documents

15 Modeling of a Voltage controlled VSI and Current Controlled VSI