Sie sind auf Seite 1von 6

A Comparative Study of Central and Distributed

MPPT Architectures for Megawatt Utility and Large


Scale Commercial Photovoltaic Plants
Ahmed Elasser, Mohammed Agamy, Juan Sabate, Robert Steigerwald, Rayette Fisher and Maja Harfman-Todorovic

General Electric Global Research Center


Niskayuna, NY, USA

Abstract- In this paper different distributed PV architectures that address this issue from a system level point of view [5,6].
are studied from an energy yield perspective. These distributed This paper evaluates the benefits of using a distributed PV
architectures are applied to massively paralleled thin film plants architecture as compared to the conventional plants with a
employing high voltage PV modules, mc-Si plants with long
central inverter based on a comparison of the energy yield of
series strings of low voltage modules and plants with medium
voltage thin film modules in order to evaluate the effectiveness each architecture, reliability and cost. The paper is organized
of the distributed architecture in each case. The effects of partial as follows: section II presents the layout of the plants under
shading, module mismatch and cable losses are quantified in study. In section III, the modeling of the PV modules and
order to obtain the energy yield for each of the architectures simulation of the effects of partial shading is presented. In
under study. The results of this trade-off study are used to section IV the energy yield calculation and the factors
quantify the benefits of a distributed architecture as well as
influencing it are presented for the different plants and
determine the optimal location of the dc/dc converters that
perform the MPPT function. module technologies. A tradeoff between the different
architectures is given in section V and finally some
concluding remarks are presented in section VI.
I. INTRODUCTION
Today’s typical photovoltaic (PV) system comprises many II. UTILITY SCALE PV ARCHITECTURES
PV modules that are connected together in series and parallel In this study, PV plants with three different module
to form strings and sub-arrays. The sub-arrays are further
technologies are evaluated:
connected using combiner boxes to feed the central inverters.
(i) High voltage tandem amorphous silicon
These PV systems become more complex as the size of the modules (Open circuit voltage ~240V)
installation increases. The success of these large-scale
(ii) Medium voltage tandem amorphous silicon
systems depends on the ability of the PV installation to
modules (Open circuit voltage ~60V)
deliver high energy yield and greater system availability [1]. (iii) Low voltage multi-crystalline silicon modules
In a central inverter architecture, all the PV modules feed into
(Open circuit voltage ~33V)
a high power central dc-dc converter stage that performs the
The number of modules connected in series per string differs
maximum power point tracking (MPPT) and then to a dc-ac
for each module technology based on the module voltage.
inverter to connect to the utility grid. This architecture does
High voltage modules are connected in a massively paralleled
not extract the maximum available energy that can be
layout with only 2 series modules per string while low
generated by the PV modules due to several factors that
voltage modules form longer strings of 15 series modules.
include: The global maximum power point (MPP) is not the
Several distributed architectures can be implemented based
MPP for all individual modules in the cases of partial
on the placement of the dc/dc converter and the MPPT
shading, soiling and mismatches leading to an energy loss,
controller associated with it. The dc/dc converter can be
power losses in the dc-distribution cables, static and dynamic
located at the central inverter, at the array combiner, string
MPPT losses and energy losses due to system outages combiner, string or module. Fig. 1 shows general plant
because of faults or maintenance. Using distributed PV
layouts for different module technologies and Fig. 2 presents
architectures provides means to alleviate these drawbacks.
the possible locations of the dc/dc converter for the different
Distributing the dc/dc converter stage such that there is one distributed plant architectures for a 1.34 MWp plant
converter per sub-array, or one converter per string or
implemented with each of the module technologies under
ultimately one converter per module can provide greater
study.
flexibility, modularity and more accurate maximum power
extraction as well as the ability to have a regulated high dc
voltage that gives lower cable losses. [2-4]
While there are a number of studies and papers published
in the literature about distributed architectures, there are few

978-1-4244-5226-2/10/$26.00 ©2010 IEEE 2753


50 Strings 50 Strings

Array Combiner

12 String
Combiners

Inverter

(a) Fig. 2 The different distributed architectures of a PV plant: (a) central dc-dc
converter, (b) Array combiner dc-dc converters, (c) String Combiner dc-dc
converters, (d) String dc-dc converters and (e) Module dc-dc converters

String
modules as shown in Fig. 4. These models are then used to
Combiner derive transfer functions relating the maximum power (Pmax)
of the PV modules to the irradiance (G) and the ambient
temperature (Ta).
8 String
Combiners
Pmax = f (G, Ta ) (1)
112 Strings
Fig. 5 shows a graphical representation of the relation
Pmax=f(G,Ta) for the two module technologies under study.
112 Strings
8 String
Combiners

String
Combiner

Fig. 3 Single diode model used to simulate the PV module.

(b)
Fig. 1 The plant architectures for the 1 MW PV plant under study The next step is studying the effect of partial shading.
(a) with high voltage thin film modules, (b) with low voltage mc-Si Shading happens in many ways: for residential and
modules commercial installations, shading happens because of
physical obstacles such as trees, chimneys, poles, other
III. MODULE MODELING AND PARTIAL SHADING buildings, clouds, etc.. For utility scale and ground mount
SIMULATION installations, shading happens mostly because of cloud
patterns and in some cases because of weeds and grass that
The first step to evaluate the energy yield is modeling the
overgrows or obstacles that are not easily removed. When
PV modules. A single diode model, shown in Fig. 3, is used
shading happens, the PV module performance is affected
to simulate the PV modules. The models show good
because of significant reduction in the irradiance, hence the
agreement with datasheet characteristics of the
output current. In string designs, when one module is affected

2754
9
G=1000W/m2, Tcell=25oC
by shading, it impacts the whole string unless it is bypassed
8 by a bypass diode. Bypass diodes tend to overheat and
7 increase system losses [2,3,7-10]. Different partial shading
scenarios are applied to the PV plants under study and the
Module Current (A)

6
maximum extracted power for each distributed architecture
5
G=800W/m2, Tcell=45oC shown in Fig. 2 with the MPPT control being implemented at
4
the dc-dc converter is evaluated. Fig. 6 shows examples of
3 different partial shading scenarios studied that in clued
2 uniform, non-uniform shading patterns and conditions where
G=1000W/m2, Tcell=60oC
1 some of the strings are fully shaded.
0
0 5 10 15 20 25 30 35
Bias Voltage (V) Partially shaded
strings
4

3.5
G=1000W/m2
3 Fully shaded
Module Current (A)

G=800W/m2
2.5
strings

2 G=600W/m2
(a)
1.5 G=400W/m2

1
G=200W/m2
0.5

0
0 00
0
10
10 00 2000
20 3030
00 4000
40 500 00 6000
60 70
0 00
Bias Voltage (V)
Fig. 4 Module model results (solid lines) vs. datasheet characteristics
(dotted lines) for (a) mc-Si modules and (b) amorphous Si tandem
modules
250

(b)
Maximum Module Power (W)

200
Fig. 6 Examples of partial shading scenarios (a) for massively
paralleled plants, (b) for plants with multiple series connected modules
150 Pmax @ Ta=55C
Pmax @ Ta=25C
Depending on the shading pattern and plant architecture, the
100 Pmax @ Ta=-55C
percentage of generated power that can be extracted from the
50 PV plant can be determined. Placing the MPPT control at
each module maximizes the power extraction and this starts
0 decreasing as we move the MPPT closer to the central
0 200 400 600 800 1000 inverter. Fig. 7 shows some comparisons between the
Irradiance (W/m2)
increase in maximum extracted power for the different
(a)
architectures with different MPP tracker locations. It is worth
160
mentioning that for systems with longer strings of series
Maximum Module Power (W)

140
modules, partial shading has a more significant impact on the
120
total generated power as compared to massively paralleled
100 Pmax @ Ta=55C
Pmax @ Ta=25C
plants.
80
60
Pmax @ Ta=-55C IV. ENERGY YIELD EVALUATION
40
20
Energy yield is the key metric in our comparison of the
central and distributed architectures. The increase in energy
0
yield will determine if the distributed MPPT approach is
0 200 400 600 800 1000
justifiable or not. Energy yield is affected by a number of
Irradiance (W/m2) factors, chief among them are: Irradiance- The total
(b) irradiance received by a plant over a year in a given location
Fig. 5 Transfer function Pmax=f(G, Ta) for (a) mc-Si modules and (b) determines the maximum energy produced.
thin film amorphous Si tandem modules

2755
45% in the previous section and the power is integrated over the
40% whole year for different architectures. Partial shading due to
power 35% clouds is represented by comparing the actual irradiance to a
moving average as shown in Fig. 8.
Percent increase in output
30%
AC to CI
p

25% SC o CI
20% String to CI
Module to CI
15% 250

10% 200

Irr-Irr_av
5% 150
0%
100
10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Percent of Plant Shaded 50
(a)
0
70
1-Jan 31-Jan 2-Mar 1-Apr 1-May 31-May 30-Jun 30-Jul 29-Aug 28-Sep 28-Oct 27-Nov 27-Dec
Date
Percent increase in output power

60

50 1000 1000
900 900
3-MPPT
40
800 800
SC to CI 700 700
600

W/m2
600

W/m2
Str ing to CI 500
30 500
400 400
Module to CI 300 300
20 200 200
100 100
10 0 0
1 6 11 16 21 1 6 11 16 21
Hour Hour
0
12.50% 25% 37.50% 50% 62.50% 75% 87.50%
Percent of Plant Shaded Fig. 8 An actual shading pattern over one year with daily moving average
(b)
Fig. 7 Example of the increase in output power gained by distributing the
MPP trackers under partial shading. Shaded panels are at G=250W/m2 and
The annual energy yield of a 1.34 MWp PV plant for two
unshaded panels are at 1000W/m2 for: (a) Massively paralleled high different geographic locations (Phoenix, AZ and Newark, NJ)
voltage thin-film plant, (b) Medium voltage thin film plant with 8 series for low voltage silicon and a high voltage thin film modules
modules per string and with different MPPT locations are summarized in Tables
(AC: Array Combiner, CI: Central Inverter, SC: String Combiner)
I and II respectively. These results show an increase in energy
yield obtained due to distributing the MPPT controllers as
Ambient temperature: The ambient temperature affects he compared to central inverter systems.
cell temperature and subsequently the module power. mc-Si
have a higher temperature degradation coefficient than thin B. Cable Losses in DC Distribution Network
film plants. Module type: The module technology plays an
important role in the total amount of energy produced. Higher Distributed systems with local dc/dc converters can regulate
efficiency modules lead to smaller footprint and higher and maintain a high voltage dc-bus, therefore reducing the
output; low efficiency modules (Thin films) require larger current in the dc-cables by the square of the voltage ratio and
area but will also lead to higher output (better temperature improving the dc/ac stage efficiency, hence improving the
coefficient and ability to convert direct an diffuse light). overall energy yield. Tables III, IV and V show the gain in
String arrangement: Depending on the number of modules energy yield for low voltage mc-silicon module plant, a
required to form a string, higher the number, higher are the medium voltage a-Si plant and high voltage thin film
losses due to partial shading, with fewer modules per string, amorphous silicon plant of a 1.34MWp rating, respectively at
the shading losses are reduced. Shading and soiling: This is a two different (575V and 750V) dc bus voltages. The baseline
major factor in how energy yield is affected. Partial, full dc-bus voltage is 360V for thin film modules and 390V for
shading, soiling all lead to a decrease in module output, hence mc-Si modules.
a reduction in energy output. Module mismatch: Mismatch
between modules leads to plant operating outside their C. Module Mismatch
optimal point, hence a reduction in energy yield. Dc-wiring In order to study the effect of module mismatch on the energy
network: the losses in the dc-cables are affected by yield for central and distributed plant architectures, a 10%
distributing the dc-dc converters and adjusting the dc-bus
voltage level [11-13]. In the following subsections the tolerance is applied to the PV module parameters along with a
approach used to calculate the effect of the different factors normal distribution function for the probability of the
on the energy yield is briefly explained. occurrence of this mismatch [14]. This analysis shows
0.5%~1% gain in energy yield for distributed systems.
A. Environmental Conditions (Irradiance and Temperature)

To compute the energy yield, irradiance and temperature


profiles for a full year are fed to the transfer functions derived

2756
TABLE I
Annual energy yield increase from partial shading for low voltage Si module plant
MPPT CENTRAL STRING STRING MODULE
LOCATION DC/DC COMBINER
CONVERTER
ENERGY 2,675,055 2,681,674 2,692,368 2,698,570 PHOENIX,
(kWh) AZ
1,987,168 1,995,769 2,008,975 2,016,993 NEWARK,
NJ
% YIELD - 0.25% 0.65% 0.88% PHOENIX,
INCREASE AZ
- 0.43% 1.10% 1.50% NEWARK,
NJ
TABLE II
Annual energy yield increase from partial shading for high voltage thin film module plant
MPPT CENTRAL ARRAY STRING STRING MODULE
LOCATION DC/DC COMBINER COMBINER
CONVERTER
ENERGY 2,328,391 2,336,349 2,338,915 2,338,915 2,341,024 PHOENIX,
(kWh) AZ
1,704,986 1,715,765 1,719,167 1,719,167 1,722,152 NEWARK,
NJ
% YIELD - 0.34% 0.45% 0.45% 0.54% PHOENIX,
INCREASE AZ
- 0.63% 0.83% 0.83% 1.01% NEWARK,
NJ

TABLE III
Energy yield gain due to cable loss reduction by increasing the dc-bus down into constant losses and load dependent losses, it is
voltage in a mc-Si plant expected that using low power converters leads to a drop in
Vdc Central dc/dc String combiner String the system efficiency and a decrease in energy yield. The
converter drop in energy yield due to lower power conversion
390V - efficiency is estimated to range from 0.2% at string combiner
(Baseline)
575V 0.06% 0.435%
level to 1.9% for module power converters.
750V 0.11% 0.585% E. Other factors Affecting Energy Yield in Distributed
Systems
TABLE IV
Energy yield gain due to cable loss reduction by increasing the dc-bus One of the key advantages of having a distributed plant
voltage in a medium voltage a-Si plant
architecture is the increased availability of power plant, since
Vdc Central dc/dc String combiner String a failure in a single dc-dc converter or the disconnection or
converter malfunction of a group of strings does not lead to a complete
360V - plant shutdown.
(Baseline)
575V 0.222% 0.228% Furthermore, other effects like soiling or accumulation of
750V 0.281% 0.288%
snow on PV modules act in a similar way to partial shading
and distributed systems can allow for up to 1% more energy
Table V yield under such conditions.
Energy yield gain due to cable loss reduction by increasing the dc-bus Another factor affecting energy yield is the MPPT control
voltage in a high voltage a-Si plant algorithm, which can be made most efficient at a module
Vdc Central dc/dc String combiner String level converter, but the closer it is moved towards the central
converter inverter the less efficient it becomes due to static and
360V - dynamic losses introduced by the MPPT controller. MPPT
(Baseline) can account for up to 2% of extra energy gain by moving it to
575V 0.058% 0.13%
a module converter.
750V 0.076% 0.166% Adding the different factors that affect the energy yield, a
gain in the range of 4~12% can be obtained for distributed
D. Converter Efficiency systems over central inverter systems. This range is
dependent on the geographic location of the plant which has a
While distributing dc-dc converters provides more control great effect on its operation.
flexibility and better MPP tracking, maintaining very high
efficiency for small power converters becomes more
challenging. Since converter losses can generally be broken

2757
V. ARCHITECTURE TRADE-OFF [8] R. Bruendlinger, B. Bletterie, M. Milde & H. Oldenkamp, “Maximum
Power Point Tracking Performance Under Partially Shaded PV Array
Conditions,” Proceedings of the 21st European PVSEC, Dresden, 2006.
Combining the different factors under study that affect the [9] R. Ramabadran & B. Mathur, “Effect of Shading on Series and Parallel
energy yield: partial shading, soiling, module mismatch, cable Connected Solar PV Modules”. Modern Applied Science, Vol. 3, N010,
losses, MPPT losses and system availability a comparison October 2009, pp. 32-41.
between the centralized PV plant and the different distributed [10] M. Garcia, J. Maruri, L. Marroyo, E. Lorenzo & M. Perez,”Partial
architectures can be performed. A complete comparison will Shadowing, MPPT Performance and Inverter Configurations:
require cost, net present value (NPV) and levelized cost of Observations at Tracking PV Plants,” PROGRESS IN
PHOTOVOLTAICS: RESEARCH AND APPLICATIONS, April 2008,
electricity (LCOE) data. A selected distributed architecture pp. 529-536.
will need to generate enough energy yield needed to [11] U. Schwabe & P. Jansson, “Performance Measurement of Amorphous
compensate for the increased cost in each architecture. While and Monocrystalline Silicon PV Modules in Eastern U.S. Energy
a detailed cost analysis have not been done for this production versus ambient and module temperature,” I2MTC 2009 -
architecture, judging from the increased energy yield obtained International Instrumentation and Measurement Technology Conference,
from the various factors listed above, distributed architectures Singapore, 5-7 May 2009
can be cost competitive at the string or string combiner level [12] S. Krauter, A. Preiss, N. Ferretti &P. Grunow “PV Yield Prediction For
for large commercial (>100kW) and utility scale (>500kW) Thin Film Technologies and the effect of input parameters inaccuracies,
23rd European Photovoltaic Solar Energy Conference, Valencia, Spain,
plants. For small commercial and residential systems, a 1–5 September 2008.
module level MPPT has been proved to be beneficial and to [13] http://www.steveransome.com/pubs.html
provide enough energy yield increase as well as other benefits [14] A. Chouder, S. Silvestre, “Analysis Model of Mismatch Power Losses in
(e.g. Module monitoring, easy installation…) [15]. PV Systems,” Journal of Solar Engineering, May 2009, Vol. 131
[15] http://enphaseenergy.com/support/learningcenter.cfm
VI. CONCLUSION
This paper presents a tradeoff study of different distributed
utility scale and large-scale commercial PV power plants. The
distribution of dc/dc converters along with the MPPT
controllers associated with them provides a substantial
increase in the annual energy yield of the system. From this
study it can be deduced that for plants with massively
paralleled modules distributing the dc/dc converters to the
array combiners provides a significantly high increase in
yield whereas, for plants with more series connected modules
significant increases in yield start to be seen at the string
combiner and string levels. As for module level distribution,
even though it provides the highest energy yield, it is
regarded to be more cost effective on commercial scale
applications rather than utility applications.
REFERENCES
[1] N. Kaushika & N. Gautam, “ Energy Yield Simulations of Interconnected
Solar PV Arrays,” IEEE Trans. on Energy Conversion, Vol. 18, No. 1,
May 2003, pp. 127-134.
[2] G. Lijun, R. Dougal, L. Shengyi & A. Lotova,“Parallel-Connected Solar
PV System to Address Partial and Rapidly Fluctuating Shadow
Conditions,” IEEE Trans. on Industrial Electronics, Vol. 56, No. 5, MAY
2009, pp. 1548-1556.
[3] H. Patel & V. Agarwal, “MATLAB-Based Modeling to Study the Effects
of Partial Shading on PV Array Characteristics,” IEEE Trans. on Energy
Conversion, Vol. 23, No. 1, March 2008, pp. 302-310.
[4] A. Chouder & S. Silvestre, “Analysis Model of Mismatch Power Losses
in PV Systems,” Journal of Solar Engineering, May 2009, Vol. 131.
[5] N. Femia, G. Lisi, G. Petrone, G. Spagnuolo & M. Vitteli, “ Distributed
Maximum Power Point Tracking of Photovoltaic Arrays: Novel
Approach and System Analysis,” IEEE Trans. on Industrial Electronics,
Vol. 55, No. 7, July 2008, pp. 2610-2621.
[6] Y. Xue, L. Chang, S. Kjaer, J. Bordonau & T. Shimizu, “ Topologies of
Single-Phase Inverters for Small Distributed Power Generators: An
Overview,” IEEE Trans. on Power Electronics, Vol. 19, No. 5, Sept.
2004
[7] J. Young-Hyok, K. Jun-Gu, P. Sang-Hoon, K. Jae-Hyung & W. Chung-
Yuen, “C-language Based PV Array Simulation Technique Considering
Effects of Partial Shading,” International Conference on, Industrial
Technology(ICIT) , 2009, pp. 1-6.

2758

Das könnte Ihnen auch gefallen