Você está na página 1de 4

GRID-TIED PV SYSTEM MODELLING: HOW AND WHY

Miles C. Russell
Ascension Technology, Inc.
P.O. Box 314, Lincoln Center, Massachusetts 01773 USA

ABSTRACT

Table 1
PV Svstem ComDonents

A critical part of Ascension Technology's operation and


monitoring of numerous grid-tied photovoltaic (PV) systems
for the U S . Environmental ProtectionAgency's PV DemandSide-Management Demonstration Projects is to insure that
these PV system are operating properly and that problems
can be diagnosed by scrutiny of the data being retrieved
from these systems. The objective of the work reported
here was to develop accurate, flexible PV array and inverter
models, and apply them to help understand the performance
of the PV systems and address issues which are important
to inverter designers/manufacturers,systems designers and
system operators.
Graphical display of measured PV system
performance, along with modelled performance has proven
useful for remotely diagnosing a wide range of PV system
phenomena and quantifying their impact: partial array
shading, partial array snow cover, PV array max-power
voltage lying outside the inverter tracking window (both an
array sizing concern, as well as an inverter design issue)
and inverter power limiting. Modelling was also key to
discovering and understanding an inverter control problem.

THE PV SYSTEMS
Ascension Technology is under contract to the U.S.
Environmental Protection Agency (EPA) to investigate the
environmental and demand-side-management (DSM)
impacts of grid-tied PV systems. During 1993, for the first
EPA PV-DSM Project, we worked with 10 electric utilities to
install individual and grouped 4-kW roof-mounted PV
systems in 17 sites around the country. The second EPA
project involves 11 electric utilities and 11 additional sites
where 18-kW PV systems (groups of three independent 6kW systems) are being roof-mounted during 1994. All of
these PV installations are equipped with instrumentation,
providing the data which supports our studies for the EPA.
Table I below identifies the components in the PV
systems for EPA projects 1 and 2. Note that these systems
are utility-interactiveand therefore do not contain batteries.
The majority of the 28 PV arrays are mounted on flat-roof
commercial buildings; four of the EPA 1 systems are
installed on residential rooftops. The arrays are fixed in tilt
and azimuth.

EPA 1

EPA 2

Siemens Solar
M55

ASE Americas
GP-50-DG

Array
Configuration

28 series x
3 parallel

10 series x
3 parallel

Array rating
(STC)

4.452 kW dc

8.55 kW dc

15 degrees

25 degrees

fixed

fixed

Omnion 4kW

Omnion 6kW

Modules

Tilt angle
Inverter
I

THE DATABASE
Many efforts to model PV systems and components
have been undertaken in the past, yet no existing software
tool provided precisely what was required for the objectives
we have with our PV system network. In addition, accurate
algorithms for translating PV current and voltage with
changes in irradiance and temperature are still being
evolved. Finally, obtaining the necessary characterization
information for PV modules has also been difficult in the
past. As a result, we have worked with Sandia National
Laboratory to obtain module characterization and models
and we have developed custom software to work with the
data being retrieved and archived from these PV
installations. Table 2 shows the parameters being
monitored at each of the EPA PV-DSM Project PV systems.
In addition, a few of the PV systems have been
equipped with extra instrumentation to monitor PV array
current and voltage. This dc-side information allows greater
scrutiny of the operation of the inverters and has proven
useful for the development of models.
The
data acquisition systems utilize Campbell
Scientific CRIO dataloggers and Ascension Technology's
Rotating Shadowband Pyranometers.

1040
CH3365-4/94/0000-1040$4.00 0 1994 IEEE

First WCPEC; Dec. 5-9, 1994; Hawaii

AC kWh produced
by the PV system

pulse-initiatingkWh meter

AC kWh used in the


host building

LiCor 200SZ pyranometer

Direct normal
irradiance

Rotating Shadowband
Pyranometer

pulse-initiatingkWh meter

Plane-of-array
irradiance

Global horizontal
irradiance

T, (cell temperature) and Tair (ambient air temperature) are


in degrees C, POA (plane-of-arrayirradiance) is in Watts per
square meter, VOC(open-circuitvoltage) and Vmp (max-power
voltage) are in volts dc, and Isc (short-circuit current) and Imp
(max-power current) are in amps dc. PV array dc power
output is computed as the product of Imp and Vmp.
These equations, with the specific constants derived for
Siemens Solar M55 and ASE Americas GP-50-DG PV
modules are used in our simulation software. In addition, we
model the efficiency of the power conditioner as a function of
the operating power level, expressed as its "percent full load
(Pfl)" -- array dc power divided by the inverter dc rating
(computed as the inverter ac rating divided by inverter full-load
efficiency). Equations of the following form are used:

Rotating Shadowband
Pyranometer

Horizontal diffuse
irradiance

Rotating Shadowband
Pyranometer

Ambient air
temperature

Thermistor housed in gilled


radiation shield

Linear region (Pfl at or below 10%)

Eff = C,*Pfl - C,

(6)

Cubic region (Pfl greater than 10%)


Eff = C, + C,*Pfl + C,*Pf12 + C,*Pf13

(7)

GFUPHICAL DATA ANALYSIS


When these models are used with the meteorological
database from Ascension Technology's growing network of
solar resource monitoring stations, PV system performance
can be predicted and system design tradeoffs made for sites
around the country. Graphical display of PV system
performance, along with modelled performance has proven
useful for remotely diagnosing a wide range of PV system
phenomena and quantifyingtheir impact. Examples are cited
below.

THE MODELS
Sandia National Laboratories has provided accurate
measurement of PV module characteristicsfor our modelling
purposes (with the permission of the module manufacturers).
The Sandia PV Systems Evaluation Laboratory suggests PV
models of the following forms. Note that constants (C,, C, et.
al.) in the equations are determined from curve fitting
measured data and will differ for each modelled parameter.

lSc = eAIC,+C,(ln(POA)-C,}] + C,(POA/I 000)(TC-50)

(1)

Imp= e"[C,+C,{ln(POA)-C,}]

(2)

+ C,(POAAOOO)(T,-50)

V, =C1+C2*In(POA/1000) - C,(T,-25)
,V
,

=C,-C, * In(POA/lOOO) - C,(T,-25)

T, = Tair

C,

C,(POA),

(3)
(4)
(5)

Array-to-Inverter Power Matching


Observe in Table 1 that the PV array output for the Project
1 PV systems (8.55 kWdc each, or 25.6 kWdc total) exceeds
the rating of the inverter (6 kWac each, or 18 kWac total).
Operationally this results in "clipping" of the PV array output
power when the inverter reaches its power limit. Figure 1
shows an example of a single day's performance for a PV
system located in Lawton, OK. The figure shows the
measured ac power output of the system, along with the
simulated output. The system reaches its nominal ac peak
output of 18 kWac and maintains this power level for several
hours around solar noon. In order to continue operating at its
limit, the inverter moves the dc votlage above the PV array's
maximum power point, until a point is found where array
current, voltage and power are all within acceptable range.
The simulated ac power output, also shown in the figure, is
not restricted by a power limit. That is, the simulated output
shows that if the inverter were able to accommodate it, the
system could produce higher power during the period of
power limiting behavior. The obvious question is: how much
energy is lost over the year due to this somewhat-oversized
PV array? With our PV system models and meteorological
database we are able to answer that question, as shown in
Figures 2 and 3.
PV array performance was simulated for an entire year,
using meteorological data for Tempe, AZ and Waltham, MA.

1041

The meteorologicaldata are 15-minute averages of irradiance


and ambient temperature. At each time-step the maximum
power of the PV array was computed. The data were then
sorted to produce the power duration curves shown in Figures
2 and 3. Note that a lower limit of 100 W/m2 was imposed on
the POA irradiance for the simulations, since our models are
not tested at extremely low irradiance.
The area beneath these curves is the annual dc energy
which could be produced by the system; for Tempe, Ai! this
is 18,350 kWh dc and for Waltham, MA 13,670 kWh dc. The
area under the portion of the curves at dc power levels of 6.6
kW (the array dc power when the inverter ac output is 6kW,
its limit) and higher is the energy which is "lost" when the
inverter limits power (see Fig. 1). In Tempe, Ai!,the lost
energy is 439 kWh dc, or 2.4% of the annual potential total;
in Waltham, MA the lost energy is 305 kWh, or 2.2% of the
potential total.

An inverter Control Problem


On a routine basis the data retrieved from the EPA PV
systems is summarized in tabular form.
Anomalous
performance is flagged. To scrutinize further, the data are
displayed with Ascension Technology's custom graphing
software. The simulated performance is displayed along with
the measured performance; in this way, discrepancies
between actual and expected performance are clearly
identified.
Figure 4 shows an example of an inverter logic problem,
which was identified and diagnosed by Ascension
Technology's data analysis procedures. Compare the
simulated and measured PV system ac power output. Note
that the PV system began operating then levelled off
unexpectedlyat a low power level of about 750 W ac; output
did not rise to expected levels until approximately 1:30 pm.
Performance of this type was observed on several other days
also. The inverter manufacturer was notified and the
phenomenon turned out to be a control problem, which was
corrected by a simple software modification.
Note that the simulated PV system ac power aligns
reasonably well with the measured values during the
afternoon, when inverter behavior is normal. This gives us
some confidence in the accuracy of the models being used.

Future Plans
Our modelling abilities will continue to be refined, using the
database of PV system performance measurements for model
validation. W e intend to adopt an I-V curve based PV array
model, to allow better investigation of array-inverter interface
design issues.
Battery storage capacrty will be modelled and simulations of
grid-tied PV systems performed to estimate the potential
benefits of shifting PV output for better demand-sidemanagement beneft. This work is being conducted as part of
the studies for the EPA PV-DSM Projects and related activities
being conducted with Delmarva Power and Light under the
National Renewable Energy Laboratory's PV BONUS
program.

Final Comments
PV system models have proven to be invaluable for
troubleshooting hardware problems and advancing our
understandingof the PV system performance issues. Insuring
that the EPA PV systems are operating properly -- diagnosing
problems using our database -- is one of the highest priorities
of the project and a major challenge of managing these
numerous widespread PV installations. Key to meeting this
challenge is the ability to model the PV systems accurately.

1200

300

600

am

900

1200

300

600

900

Pm

lime of Day (June 13,1994)

1 BMeasured Power Output -Simulated

Power Output1

Figure 1 Simulated and measured PV system ac output,


showing inverter power limiting

inverter Maximum-Power-Point Tracking


In a similar situation, where the daily data summary had
flagged low system energy production, relative to our
expectations based on simulation, we identified an apparent
maximum-power-tracking problem. This problem arose at a
site which has dc instrumentation, in addition to the standard
irradiance and ac transducers.
Figure 5 shows the measured and simulated PV array
maximum power voltage for a Project 1 site at Edwards Air
Force Base, CA. As shown, the simulated array max-power
voltage varies over the day, dropping just below 200 Volts at
mid-day when the PV array temperature is highest; the
measured array voltage locks up at a level of approximately
220 Volts, contrary to manufacturer's specifications. This
behavior has been observed on numerous occasions. At this
time we are working with the inverter manufacturer to
understand the cause of this behavior.

1042

[racthn o

7 exceeds P

Photovoltaic Power (kWac)

25

02 . . . . . . . . . . . . . . . . . . . . . . .

izoo
00

300

600

goo

am

10

1200

600

300

goo

Pm

Time of Day (May 19,1994)

P,anaymax power (kWJ

/=Measured Power Output -Simulated Power Output]

I--Tempe,J
Figure 2 Max-power duration curve, Tempe AZ, for a
simulated EPA2 PV array, south-facing, 25-degree tilt

Figure 4 Simulated and measured PV system ac output for


an Project 2 system with an inverter maximum power tracking
problem

10

0.8

06

04

02

izoo

600

300
am

goo

1200

300

600

goo

pm

lime of day (May 4,1994)

00

10

1:SS:Measured operating voltage -Simulated maximum power voltage

F: a m y max pawer (kw)


1-Waltham, MA

Figure 3 Max-power duration curve, Waltham, MA, for a


simulated EPA2 PV array, south-facing, 25-degree tilt

Figure 5 Simulated and measured PV system ac output for


an EPA Project 1 system with inverter voltage control unable
to track the maximum power point below 220 volts

1043

Você também pode gostar