IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 2, JUNE 2011 707
Power Loss Comparison of Single- and Two-StageGrid-Connected Photovoltaic Systems
Tsai-Fu Wu, Senior Member, IEEE, Chih-Hao Chang, Li-Chiun Lin, and Chia-Ling Kuo
Abstract—This paper presents power loss comparison of single-and two-stage grid-connected photovoltaic (PV) systems based onthe loss factors of double line-frequency voltage ripple (DLFVR),fast irradiance variation + DLFVR, fast dc load variation +DLFVR, limited operating voltage range + DLFVR, and over-all loss factor combination. These loss factors will result in powerdeviation from the maximum power points. In this paper, bothsingle-stage and two-stage grid-connected PV systems are consid-ered. All of the effects on a two-stage system are insignificant dueto an additional maximum power point tracker, but the trackerwill reduce the system efficiency typically about 2.5%. The powerloss caused by these loss factors in a single-stage grid-connectedPV system is also around 2.5%; that is, a single-stage system hasthe merits of saving components and reducing cost, and does notpenalize overall system efficiency under certain operating voltageranges. Simulation results with the MATLAB software packageand experimental results have confirmed the analysis.
Index Terms—Loss factor, power loss comparison, single-stagegrid-connected photovoltaic (PV) system, two-stage grid-connectedPV system.
I. INTRODUCTION
PHOTOVOLTAIC (PV) grid-connected systems based ona two-stage configuration have been widely studied. Re-
cently, PV dc-distributed systems, as shown in Fig. 1, witheither a single-stage configuration [saving an maximum powerpoint tracker (MPPT) stage] or a two-stage one have been beingemerging [1]. They can draw maximum power from PV mod-ules and inject the power into utility grid with unity power factoror they can rectify the ac source to replenish and regulate the dcbus. However, the loss factors, such as operational conditions,components, and grid voltage, will deviate effective PV outputpower.
In a grid-connected PV system (GCPVS), PV power varieswith operational conditions, such as irradiance, temperature,light incident angle, reduction of sunlight transmittance on glassof module, and shading [2]–[5], [12], [13]. These factors havebeen investigated in detail and the authors have presented diag-nosis methods to estimate the reduction of PV power. Moreover,a special condition, snow coverage, has been also discussed [6],which was compared with other coverage situations such as
Manuscript received October 21, 2010; revised January 23, 2011; acceptedFebruary 28, 2011. Date of publication April 19, 2011; date of current versionMay 18, 2011. Paper no. TEC-00417-2010.
Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TEC.2011.2123897
Fig. 1. Block diagram of a PV dc-distributed system.
shading and dirt. Then, another loss factors, such as compo-nents, solar cell serial resistance, and capacity loss in PV bat-teries, have been reported [3], [7], [8], [19], which are relatedto the cells themselves. During system operation, there are lossfactors such as double line-frequency voltage ripple (DLFVR)due to ac grid [9], [10], [14]–[17] and fast irradiance varia-tion [11], causing deviation from the maximum power points(MPPs) and also resulting in power loss. However, these factorshave not been taken into account in power loss comparison forboth single-stage and two-stage GCPVS.
This paper presents power loss comparison according to fiveloss factors that might deviate the operating points from theMPPs. First, this paper conducts modeling of solar cells andnumerical analysis including I–V characteristics of PV modulesto derive the effect of DLFVR on both single-stage and two-stage GCPVS. Then, it performs power loss analysis due tofast irradiance variation. Moreover, a single-stage PV systemwill penalize its output power under certain operating voltageranges, while it can save an MPPT stage. In a PV dc-distributedsystem, as shown in Fig. 1, its dc-bus voltage is regulated bythe bidirectional inverter within 360–400 V (380±20 V), and dcappliance and equipment can be connected to the dc bus directly.Thus, this study chooses “fast dc load variation” and “limitedoperating voltage range” as two of the loss factors. Since theeffect of DLFVR exists all the time, this paper presents the fiveloss factors as “DLFVR,” “fast irradiance variation + DLFVR,”“fast dc load variation + DLFVR,” “limited operating voltagerange + DLFVR,” and “overall loss factor combination.”
II. MODELING OF PV ARRAYS
This section models a PV array, consisting of 14 in seriesand 2 in parallel PV modules (APOS Series AP-220) for a
708 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 2, JUNE 2011
Fig. 2. Equivalent circuit of a solar cell.
single-stage GCPVS or 7 in series and 4 in parallel for a two-stage GCPVS. The power loss analysis based on the loss factorswill be then presented in the next section.
An equivalent circuit of a solar cell is shown in Fig. 2, inwhich the current source Isc is the light-induced current, Rj
is the nonlinear resistance in the P–N junction, Dj is the P–Njunction diode, Rsh and Rs stand for internal equivalent paralleland series resistance, respectively, Ro is the output load, andipv and vpv are the output current and voltage of the solar cell,respectively [18]. For simplifying the analysis, Rsh , Rs , and Rj
are ignored. Therefore, the output current ipv of the solar cellcan be expressed as
ipv = npIsc − npIsat
[e((q/κT A)(vpv /ns )) − 1
](1)
where np is the number of solar cells in parallel, ns is the numberof solar cells in series, q is the electric charge (1.6×10−19 C), κ isthe Boltzmann constant (1.38×10−23 J/K◦), T is the temperatureof solar cells (in degrees Kelvin), and A is the diode ideal factor(A = 1–5). Additionally, Isat stands for the reverse saturationcurrent and can be expressed as
Isat = Irr
[T
Tr
]3
exp[qEgap
κA
(1Tr
− 1T
)]. (2)
The saturation current Isat is a function of temperature T. In(2), Tr is the reference temperature of the solar cell, Irr is thereverse saturation current at Tr , and Egap is the band gap ofthe solar cell (in electronvolts). Moreover, the light-inducedcurrent Isc changes with irradiance and temperature, which canbe expressed as
Isc = [Isso + Ki(T − Tr )]Si
1000(3)
where the current Isso is the short-circuit current at referencetemperature Tr and a specific irradiance (1000 W/m2), Ki isthe temperature coefficient of a short-circuit current, and Si isthe irradiance (W/m2). Table I lists the parameters of the APOSSeries AP-220 PV module [20], which are used in (1)–(3). Bytuning the ideal diode factor A = 1.8 and energy band gap Egap =1.4 eV, a PV module can be then simulated and compared witha real module, as shown in Fig. 3. It can be observed that thesimulated results are relatively close to the real ones, which canensure accuracy of the following analysis.
III. ANALYSIS OF POWER LOSS
This section presents power loss analysis of both single-stageand two-stage GCPVS, which includes the effects of DLFVR,fast irradiance variation + DLFVR, fast dc load variation +
TABLE ISPECIFICATIONS OF THE 220-W SOLAR MODULE (APOS SERIES AP-220)
Fig. 3. PV module I–V characteristics of (a) simulated module and (b) realone (APOS Series AP-220).
DLFVR, limited operating voltage range + DLFVR, and overallloss factor combination. The analysis is based on the simulatedPV array model shown in Figs. 2 and 3.
A. DLFVR
1) Single-Stage GCPVS: The circuit configuration of asingle-stage GCPVS is shown in Fig. 4. When the systemreaches the maximum power point Pmpp under a specific ir-radiance, its PV voltage vpv and current ipv become constantand equal to Vmpp and Impp , respectively. However, the effectof DLFVR on the deviation of PV voltage away from Vmpp willcause power loss, which can be expressed as
PASloss =
∑nk=1 ploss(k)
n(4)
where
ploss(k) = Pmpp − vpv(k) · ipv(k)
n =Tl
Ts.
Here, Tl is the line period and Ts is the switching period ofthe full-bridge inverter. Expressions for vpv and ipv are derived
WU et al.: POWER LOSS COMPARISON OF SINGLE- AND TWO-STAGE GRID-CONNECTED PHOTOVOLTAIC SYSTEMS 709
Fig. 4. Circuit configuration of a single-stage GCPVS.
first, and then the power loss can be determined. Voltage vpvcan be expressed as
vpv(k) = vpv(k − 1) + Δvpv(k) (5)
where
Δvpv(k) =Ts
Cpv[ipv(k − 1) − iinv (k − 1)]
vpv(0) = Vmpp
and current iinv stands for the inverter input average current.PV voltage vpv will vary with input capacitance Cpv , the dif-ference between current ipv and iinv , and switching period Ts .Additionally, current ipv will change with vpv . From (1), we canobtain a numerical expression for ipv :
ipv(k) = npIsc − npIsat
[e((q/(κT A))(vpv (k))/(ns )) − 1
](6)
where
ipv (0) = Impp .
In general, an inverter system is operated by a current controlat a fixed frequency 1/Ts . Even though the system reaches theMPP, its inductor current iL still has a current ripple; that is,the uperbound current command isd can be set to α × is for aninverter-inductor current iL to track, as shown in Fig. 5, whereα is a real number larger than unity. Moreover, the circuit shownin Fig. 4 is operated as a buck converter to shape the inductorcurrent to be sinusoidal and in phase with the ac grid voltage.Thus, the inverter input average current iinv can be approximatedby a multiplication of the output current is and the duty ratio din every switching cycle Ts , which can be expressed as
Fig. 5. Conceptual current waveforms of isd , iin , and iL .
Fig. 6. Circuit configuration of a two-stage GCPVS with a boost converterfunctioning as an MPPT.
From the aforementioned equations, we can determine theturn-on time interval as follows:
d(k)Ts =Lac [isd(k) − iL (k)]
vpv(k) − |vs(k)| (8)
where Lac is the inverter inductor and vs is the ac-grid voltage.From (1) to (8), the total power loss PAS
loss deviating from theMPP during one line period (1/60 Hz) can be estimated.
2) Two-Stage GCPVS: Circuit configuration, with a boostconverter functioning as an MPPT, of a two-stage GCPVS isshown in Fig. 6. With the boost converter, the operating voltagevpv of the PV modules can be lower than the dc-link voltagevlink . For comparing both single-stage and two-stage systems,inverter parameters are kept identical. The expression of the PVcurrent ipv is the same as (6), and the variation of the PV voltagevpv is the same as (4) except the inverter input current iinv , whichis replaced with the inductor current iLb . Determination of theboost inductor current iLb is illustrated in Fig. 7. Before the nextperturbation of maximum power point tracking, it is necessaryto regulate the average value of iLb equaling the MPP currentImpp . Therefore, we can adopt an average method to determineiLb where the average value of iLb equals Impp during turn-oninterval dbTs . It can be expressed as
iLb(k) = iLb(k − 1) + ΔiLb(k − 1)
− vpv(k − 1) − vlink(k − 1)Lb
· [1 − db(k − 1)]Ts
(9)
710 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 2, JUNE 2011
Fig. 7. Illustration of inductor current iLb varying with time.
where
ΔiLb(k − 1) = 2 (Impp − iLb(k − 1))
=vpv(k − 1)
Lb· db(k − 1) · Ts.
From (9), we can have
db(k) =2 (Impp − iLb (k)) · Lb
vpv(k) · Ts(10)
where
iLb(0) = Impp − ΔiLb(0)2
ΔiLb(0) =Vmpp
Lb· db(0) · Ts
and
db(0) = 1 − Vmpp
vlink(0).
Additionally, similar to (5), variation of the dc-link voltage vlinkcan be analogously related to the input diode current id and PVcurrent ipv . Power loss analysis according to the loss factor ofDLFVR can be attained from (6), (7), (9), and (10).
B. Fast Irradiance Variation + DLFVR
The output power of a PV array is strongly dependent on irra-diance and temperature, where irradiance would change rapidlyin a cloudy day. A cloud passing over PV modules will causeoperating point deviating from the MPP and resulting in powerloss. The power loss analysis based on fast irradiance varia-tion and DLFVR will include the relationship between PV arrayoutput current and irradiance. Then, by setting the change ofirradiance ΔSi over one line period (1/60 Hz) and combining(1)–(3) with the analysis of DLFVR, as well as the effect of fastirradiance variation on PV voltage vpv , the power loss PB
loss canbe determined as
PBloss =
∑nk=1 ploss(k)
n(11)
where
ploss(k) = Pmpp(Si(k)) − vpv(k) · ipv(k)
and
Si(k) = Si(k − 1) + ΔSi/Ts.
Fig. 8. Block diagrams of (a) single-stage and (b) two-stage GCPVS with dcload.
C. Fast DC Load Variation + DLFVR
Fig. 8 shows block diagrams of a single-stage and a two-stage GCPVS with dc load. When the dc load starts to absorbpower, the capacitor current ic will drop and the PV voltage vpvwill drop correspondingly. The PV voltage vpv then will deviatefrom the MPP and result in power loss. By setting the requiredpower Pdc for dc load and rising time Trise from no load to thefull load, the expression for dc load power pdc load is shown asfollows:
pdc load(k) = pdc load(k − 1) + Δpdc load (12)
where
Δpdc load =Pdc
Ts· Trise
and
pdc load(0) = 0.
From (12), we can obtain the dc load current
idc load(k) =pdc load(k)vlink(k)
. (13)
Analysis of power loss due to fast dc load variation can beaccomplished from (12) and (13), and based on the derivationof DLFVR.
D. Limited Operating Voltage Range + DLFVR
If a single-stage GCPVS is applied to a dc microgrid powerdistribution system, its operating voltage range of PV arrays islimited within 360–400 V. In general, an MPP voltage changeswith irradiance and temperature. Therefore, Pmpp is expressedas Pmpp (Si ,T), and it is identical to (4).
E. Overall Loss Factor Combination
Since all loss factors are not independent except the effectof DLFVR, the power loss cannot be summarized from eachindividual one. Therefore, combing all loss factors together toanalyze the total power loss is necessary. With (1)–(8), the over-all power loss due to these loss factors can be then calculated.
IV. ANALYSIS AND SIMULATION PROCEDURE
This section summarizes the procedure of the power lossanalysis and its simulation. By following the procedure, onecan conduct power loss analysis of a given PV array and aGCPV system according to the aforementioned loss factors. The
WU et al.: POWER LOSS COMPARISON OF SINGLE- AND TWO-STAGE GRID-CONNECTED PHOTOVOLTAIC SYSTEMS 711
TABLE IIOPERATING CONDITIONS OF THE TWO PV SYSTEMS
proposed modeling procedure can include both high-frequencyand low-frequency voltage ripple and power loss. Moreover, thepower loss analysis can be conducted with loss factors one byone. The procedure is described as follows.
1) Program array characteristic equations (1)–(3) into MAT-LAB and set parameters of a selected PV module, suchas open-circuit voltage Voc , short-circuit current Isc , andothers from (1) to (3) except the ideal factor A and energyband gap Egap .
2) Tune A (A = 1–5) and Egap (Egap ≈ 1.1 eV) so thatthe simulated module model can fit the actual parameters,Vmpp , Impp , and temperature coefficient Kv of the voltage.
3) Set operating conditions, such as Cpv , Clink , irradiance,and temperature, and program the derivation of power losscaused by DLFVR, (1)–(10), into MATLAB.
4) Based on step 3), set a fast irradiance change rate ΔSi
and program the derivation of power loss caused by fastirradiance variation, (11).
5) Based on step 3), set dc load conditions, such as dc loadpower Pdc and the rising time Trise , and program thederivation of power loss caused by fast dc load variation,(12) and (13).
6) Base on step 2), set ten irradiance levels from 100 to1000 W/m2 and six temperature levels from 25 to 70 ◦C toestimate the power loss caused by over limited operatingvoltage range, 360–400 V.
7) Finally, summarize the power loss due to the aforemen-tioned loss factors for both single-stage and two-stagegrid-connected PV systems.
V. SIMULATED AND MEASURED RESULTS
This section shows simulated and measured results basedon the power loss analysis, and the plots of power loss versusdifferent operating conditions, such as irradiance, dc-link ca-pacitance, and PV-side capacitance. In addition, the effects ofthe loss factors on both single-stage and two-stage GCPVS arediscussed. Table II shows the operating conditions for the twosystems.
Fig. 9. Ripple waveforms of (a) PV array voltage vpv and (b) PV array outputpower ppv in a single-stage GCPVS when capacitor Cpv is fixed at 8∗470 μF.
Fig. 10. Illustration of a power ripple due to the effect of DLFVR.
Fig. 11. Plots of PV capacitor Cpv versus (a) PV voltage-ripple percentageand (b) power-loss percentage in a single-stage GCPVS.
A. Simulated Results
1) DLFVR: For the operating conditions shown in Table II,we can obtain the waveforms of PV array voltage vpv , dc-linkvoltage vlink , and PV array output power ppv with MATLAB.The range of capacitor Cpv has six levels from 1∗470 μF to20∗470 μF in a single-stage GCPVS. The waveforms of vpvand ppv can be simulated, as shown in Fig. 9, in which thesystem is operated with Cpv = 8∗470 μF. It can be observedthat the magnitude of ppv is not constant, and its swing is muchlarger when vpv is located on the right-hand side of the MPP,which results from a steeper slope in the P–V curve, as shownin Fig. 10. Then, by calculating the voltage ripple of vpv and thepower loss, the plots of Cpv versus voltage-ripple percentageand power-loss percentage are illustrated in Fig. 11. The powerloss due to double line-frequency in a single-stage PV systemis about 3.5 W or 0.06% under Cpv = 8∗470 μF.
For a two-stage GCPVS, the dc-link capacitor Clink is fixedto 8∗470 μF, and the range of Cpv varies from 47 to 470 μF.Fig. 12 shows the ripple waveform of dc-link voltage vlink , PVarray voltage vpv , and its output power ppv with Cpv = 470 μF.The voltage ripple of vpv and the power ripple are almost zeromainly due to switching-frequency; hence, the ripple effect can
712 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 2, JUNE 2011
Fig. 12. Ripple waveforms of (a) voltages vlink and vpv and (b) PV arrayoutput power ppv in a two-stage GCPVS when a PV capacitor Clink is fixed at8∗470 μF and Cpv is 470 μF.
Fig. 13. The ripple waveforms of (a) PV array voltage and (b) PV array powerin a single-stage GCPVS when fast irradiance variation is 100 W/m2 /16.6 ms.
be ignored. Therefore, we can select PV capacitor Cpv = 470 μF,which is sufficient enough to filter out the effect of DLFVR.
2) Fast Irradiance Variation + DLFVR: First, we considerthat the speed of cloud is 30 m/s passing over PV arrays,and its corresponding irradiance variation ΔSi /Δt being about100 W/m2 /16.6 ms and the PV array temperature being un-changed over one line cycle (1/60 Hz). According to the powerloss analysis of fast irradiance variation, the ripple waveformsof PV voltage vpv and PV output power ppv can be simulatedas shown in Fig. 13. For a large PV capacitor Cpv , the changeof vpv is only 3.8 V, and the power loss is 3.67 W or 0.07%.Because the change of Vmpp during 100 W/m2 is only 2 V, theoperating voltage does not deviate from the MPP significantly.Therefore, the effect of fast irradiance variation on a single-stageGCPVS is not significant.
3) Fast DC Load Variation + DLFVR: For common dchome appliances, some like air conditioners consume highpower above 500 W per set and others like lamps and com-puters just consume low power, below 300 W. In the following,we assume that a set of dc load includes two air conditioners(the maximum power is 2000 W), ten lamps (about 400 W), twocomputers (about 500 W), and the rising times of their powerare 3 min, 1 ms, and 0.5 ms, respectively. That is, the fastestpower variation is 1 kW/ms. Moreover, the operating condi-tions, Cpv = 8∗470 μF for a single-stage GCPVS, and Clink =8∗470 μF and Cpv = 470 μF for a two-stage GCPVS, are se-lected. The simulated results can be obtained from (12) and (13)with parameters Pdc = 1000 W and Trise = 1 ms, as shownin Figs. 14 and 15, respectively. The power loss caused by adeviation from the MPP under fast dc load variation is 7.77 Won average or 0.13% in a single-stage GCPVS and 0.12 W or0.002% in a two-stage GCPVS.
Fig. 14. Ripple waveforms of (a) PV array voltage and (b) PV array outputpower under fast dc load variation in a single-stage GCPVS.
Fig. 15. Ripple waveforms of (a) voltages vlink and vpv and (b) PV arrayoutput power under fast dc load variation in a two-stage GCPVS.
Fig. 16. Plots of irradiance versus the MPP voltage under the limited operatingvoltage range of 360–400 V with 14 PV modules in series.
4) Limited Operating Voltage Range + DLFVR: As men-tioned in Section II, the operating voltage Vmpp deviating outof the limited range, 360–400 V, will cause power loss. In thefollowing, the power loss versus irradiance and temperatureunder the limited voltage range is investigated and illustrated.Note that infrequently occurring operating conditions are re-moved from the plots, since some temperature levels couldnot happen under certain irradiance, such as 25–30 ◦C underSi = 600–1000 W/m2 , 50 ◦C under Si = 100–400 W/m2 , and60 ◦C under Si = 100–500 W/m2 . With simulation, plots ofirradiance versus MPP voltage and plots of irradiance versuspower loss are illustrated in Figs. 16 and 17, respectively. Thepower loss due to the limited operating voltage range is about0–4%, and the maximum power loss under 600 W/m2 and 60 ◦Cis 3.84%.
It is important to ensure that the MPP voltage range of aPV array can fit the limited operating voltage range, whichcan prevent deviation from the MPPs and reduce the powerloss. According to the P–V characteristics of the aforementioned
WU et al.: POWER LOSS COMPARISON OF SINGLE- AND TWO-STAGE GRID-CONNECTED PHOTOVOLTAIC SYSTEMS 713
Fig. 17. Plots of irradiance versus power loss under the limited operatingvoltage range of 360–400 V with 14 PV modules in series.
Fig. 18. Conceptual P–V curve for illustrating a limited operating voltagerange under typical operational conditions.
Fig. 19. Plots of irradiance versus power loss under the limited operatingvoltage range of 360–400 V with 15 PV modules in series.
arrays, the maximum value of the MPP voltages appears at Si =600 W/m2 and T = 25 ◦C, and the minimum one appears at Si =600 W/m2 and T = 60 ◦C, as shown in Fig. 18. The number ofPV modules in series and in parallel should be chosen to ensurea proper voltage range, which is located within ±20 V or 5%variation from the maximum to the minimum value. As a result,the MPP voltage range under typical operational conditions, T =30–50 ◦C and Si = 400–1000 W/m2 , can almost fit the limitedrange.
For elaborating on this setting, we add or reduce one PVmodule from the original PV array. The added simulation resultsfor 15 and 13 PV modules in series are shown in Figs. 19and 20, respectively. It is clear that larger power loss occurs athigher irradiance for 15 PV modules in series, and the same casehappens at lower irradiance for 13 in series; that is, more or lessconnected PV modules will cause major MPP range out of theoperating voltage range, 360–400 V, significantly. Therefore, itis important to set the series number of a PV array to ensure thatthe MPP range can fit the limited operating voltage range.
5) Overall Loss Factor Combination: Since all of the lossfactors are not independent, combing them together to calcu-late overall power loss is necessary. Moreover, this analysis
Fig. 20. Plots of irradiance versus power loss under the limited operatingvoltage range of 360–400 V with 13 PV modules in series.
Fig. 21. Ripple waveforms of (a) PV array voltage and (b) PV array outputpower under overall loss factor combination in a single-stage GCPVS.
just focuses on a single-stage PV system, because the effectsdue to “fast irradiance variation + DLFVR” and “fast dc loadvariation + DLFVR” are very insignificant in a two-stage PVsystem and the effect due to “limited operating voltage range+ DLFVR” does not exist in the two-stage case. Fig. 21 showsthe PV array output voltage and power under overall loss fac-tor combination and with the operational parameters of Si =600 W/m2 , T = 60 ◦C, Pmpp = 3028 W, Vmpp = 336 V, ΔSi =−100 W/m2 /16.6 ms, Pdc = 1000 W, Trise = 1 ms, and theoperating voltage fixed to 360 V for a single-stage GCPVS. ThePV array voltage has double line-frequency ripple and drops dueto the effect of fast irradiance and fast dc load variations. Theaverage power loss over one line period is 77.57 W or 2.34%.
B. Efficiency Comparison
In a two-stage GCPVS, the loss factor, limited operating volt-age range, does not result in power loss, since a PV array is notdirectly connected to a grid-connected inverter. Moreover, anadditional boost converter with the MPPT function can updateits tracking command immediately so that the effect of fast ir-radiance variation can be ignored. As a result, the power lossesdue to these two factors are set to 0%. From the aforementionedanalysis, the loss factor of the limited operating voltage rangewill deviate the output power from the MPP most significantlyin a single-stage GCPVS. This study considers a typical boostconverter loss of about 2.5%. Then, according to the aforemen-tioned simulation, the power loss due to the loss factors underthe discussed operating conditions is shown in Table III. It canbe seen that the total loss in a single-stage GCPVS is slightlysmaller than that in a two-stage GCPVS. If the operating volt-age range of a single-stage system is properly limited, its overallefficiency will be higher than that of a two-stage system.
714 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 26, NO. 2, JUNE 2011
TABLE IIIPOWER LOSS OF THE TWO PV SYSTEMS (SIMULATED)
Fig. 22. Photograph of the prototype of single-stage and two-stage PVsystems.
C. Measured Results
In reality, it is difficult to measure the power loss due to eachsingle loss factor at a time. In this study, we verify the feasibilityof the systems with overall output power measurement. Fig. 22shows a photograph of the prototype of the single-stage and two-stage systems. When the single-stage system is operated, theboost stage is removed from the hardware set-up. Table IV showsthe measured output power of the two systems under the effectof all possible loss factors together. It can be observed that thesingle-stage system can yield higher output power. Moreover,the right two columns show the% deviations of Pout,1 and Pout,2between the simulated and measured results. It can be observedthat the deviation is less than ± 3%.
VI. CONCLUSION
This paper has presented numerical analysis with MATLABto simulate the power loss caused by deviation from the MPPswhich are due to the loss factors of DLFVR, fast irradiancevariation + DLFVR, fast dc load variation + DLFVR, limited
TABLE IVINVERTER OUTPUT POWER MEASUREMENT FROM THE TWO 6-kW PV SYSTEMS
operating voltage range + DLFVR, and overall combination forboth single-stage and two-stage GCPVS. Analysis and simu-lation procedure of the power loss has been described, whichprovides engineers a guideline for building a model of PV arrayand performing the power loss analysis. According to the lossanalysis, the total power loss in a single-stage GCPVS is closeto a two-stage GCPVS, while the single-stage one can save astage of a boost converter. That is, from a viewpoint of effi-ciency, cost, and system size, a single-stage GCPVS is feasiblein dc-distribution and grid-connected applications if the operat-ing voltage range is properly selected. It has been also verifiedwith the measured inverter output power from the two systems.
REFERENCES
[1] T.-F. Wu, C.-H. Chang, Y.-D. Chang, and K.-Y. Lee, “Power loss analysisof grid-connection photovoltaic systems,” in Proc. Power Electronics andDrive Systems (PEDS) Conf., 2009, pp. 326–331.
[2] T. A. Huld, M. Suri, R. P. Kenny, and E. D. Dunlop, “Estimating PV perfor-mance over large geographical regions,” in Proc. 31st IEEE. Photovoltaic.Spec. Conf., 2005, pp. 1679–1682.
[3] N. Okada, S. Yamanaka, H. Kawamura, and H. Ohno, “Diagnostic methodof performance of a PV module with estimated power output in consideringfour loss factors,” in Proc. 31st IEEE Photovoltaic. Spec. Conf., 2005,pp. 1643–1646.
[4] R. Suzuki, H. Kawamura, S. Yamanaka, H. Ohno, and K. Naito, “Lossfactors affecting power generation efficiency of a PV module,” in Proc.29th IEEE. Photovoltaic. Spec. Conf., 2002, pp. 1557–1560.
[5] N. Okada, S. Yamanaka, H. Kawamura, and H. Ohno, “Energy loss ofphotovoltaic system caused by irradiance and incident angle,” in Proc.3rd IEEE Photovoltaic Energy Convers. World. Conf., 2003, pp. 2062–2065.
[6] Y. Ueda, K. Kurokawa, T. Itou, K. Kitamura, Y. Miyamoto, M. Yokota,and H. Sugihara, “Performance ratio and yield analysis of grid connectedclustered PV systems in Japan,” in Proc. 4th IEEE Photovoltaic EnergyConvers. World. Conf., 2006, pp. 2296–2299.
WU et al.: POWER LOSS COMPARISON OF SINGLE- AND TWO-STAGE GRID-CONNECTED PHOTOVOLTAIC SYSTEMS 715
[7] W.-M. Wu, X.-L. Wang, P. Geng, and T.-H. Tang, “Efficiency analysisfor three phase grid-tied inverter,” in Proc. IEEE Ind. Technol. Int. Conf.(ICIT), 2008, pp. 1–5.
[8] T. Oozeki, K. Otani, and K. Kurokawa, “An evaluation method for PVsystem to identify system losses by means of utilizing monitoring data,”in Proc. 4th IEEE Photovoltaic Energy Convers. World Conf., 2006,pp. 2319–2322.
[9] N. A. Ninad and L. A. C. Lopes, “Operation of single-phase grid-connectedinverters with large DC bus voltage ripple,” in Proc. IEEE Energy Powerand Control (EPC), 2007, pp. 172–176.
[10] G. Ertasgin, D. M. Whaley, N. Ertugrul, and W. L. Soong, “Analysisand design of energy storage for current-source 1-ph grid-connected PVinverters,” in Proc. 23rd Annu. IEEE Appl. Power Electron. Conf. (APEC),2008, pp. 1229–1234.
[11] D. D. Nguyen and B. Lehman, “Modeling and simulation of solar PVarrays under changing illumination conditions,” in Proc. IEEE Comput.Power Electronics (COMPEL’06) Workshops, 2006, pp. 295–299.
[12] H. Patel and V. Agarwal, “MATLAB-based modeling to study the effectsof partial shading on PV array characteristics,” IEEE Trans. EnergyConvers., vol. 23, no. 1, pp. 302–310, Mar. 2008.
[13] X. Weidong, N. Ozog, and W. G. Dunford, “Topology Study of photo-voltaic interface for maximum power point tracking,” IEEE Trans. Ind.Electron., vol. 54, no. 3, pp. 1696–1704, Jun. 2007.
[14] D. Casadei, G. Grandi, and C. Rossi, “Single-phase single-stage photo-voltaic generation system based on a ripple correlation control maximumpower point tracking,” IEEE Energy Convers., vol. 21, no. 2, pp. 562–568,Jun. 2006.
[15] A. Kotsopoulos, J. L. Duarte, and M. A. M. Hendrix, “Predictive DCcontrol of single-phase PV inverters with small DC link capacitance,” inProc. IEEE Int. Symp. Ind. Electron., Jun. 9–11, 2003, vol. 2, pp. 793–797.
[16] N. D. Benavides and P. L. Chapman, “Modeling the effect of voltageripple on the power output of photovoltaic modules,” IEEE Trans. Ind.Electron., vol. 55, no. 7, pp. 2638–2643, Jul. 2008.
[17] S. B. Kjaer, J. K. Pedersen, and F. Blaabjerg, “A review of single-phasegrid-connected inverters for photovoltaic modules,” IEEE Trans. Ind.Appl., vol. 41, no. 5, pp. 1292–1306, Sep./Oct. 2005.
[18] S. Liu and R. A. Dougal, “Dynamic multiphysics model for solar array,”IEEE Trans. Energy Convers., vol. 17, no. 2, pp. 285–294, Jun. 2002.
[19] T. Hund, “Capacity loss in PV batteries and recovery procedures,” Pho-tovoltaic System Applications Department, Sandia National Labs website(1998). [Online]. Available: http://www.sandia.gov/pv/docs/PDF/caploss.pdf
Tsai-Fu Wu (S’88–M’91–SM’98) received the B.S.degree in electronic engineering from NationalChiao-Tung University, Hsinchu, Taiwan, in 1983,the M.S. degree in electrical and computer engineer-ing from Ohio University, Athens, Greece, in 1988,and the Ph.D. degree in electrical engineering andcomputer science from the University of Illinois,Chicago, in 1992.
From 1985 to 1986, he was a System Engineerat SAMPO, Inc., Taiwan, where he was involved indeveloping and designing graphic terminals. From
1988 to 1992, he was a teaching and research assistant in the Department ofElectrical Engineering and Computer Science (EECS), University of Illinois.Since 1993, he has been with the Department of Electrical Engineering, Na-tional Chung Cheng University, Chia-Yi, Taiwan, where he is currently a ChairProfessor and Director of the Elegant Power Application Research Center. Hisresearch interests include developing and modeling of power converters, designof electronic dimming ballasts for fluorescent lamps and metal halide lamps,design and development of smart green energy dc-distribution systems with gridconnection.
Dr. Wu received three Best Paper Awards from Taipei Power Electronics As-sociation in 2003–2005. In 2006, he was awarded as an outstanding researcherby the National Science Council, Taiwan.
Chih-Hao Chang is working toward the Ph.D. de-gree in Elegant Power Application Research Center,National Chung Cheng University, Chia-Yi, Taiwan.
His research interests include three-phase grid-connected inverter, three-phase power factor correc-tion, and dc microgrid.
Li-Chiun Lin was born in Taiwan in 1986. He is aMaster’s student in Elegant Power Application Re-search Center, National Chung Cheng University,Chia-Yi, Taiwan.
His research interests include grid-connected in-verter, power factor correction, DSP-based controland dc-microgrid distribution system.
Chia-Ling Kuo was born in Taiwan in 1985. Shereceived the B.S. degrees in 2008 and the M.S. de-gree in 2010 in electrical engineering from NationalChung Cheng University, Chia-Yi, Taiwan, where sheis currently working toward the Ph.D. degree in thesame department.
Her research interests include design and im-plementation of bidirectional inverters for dc-distribution applications.
