Production Modelling Grid-Tied PV Systems

Production Mo
for Grid-Tied PV Systems

Production modeling meets multiple needs. Integrators seek to optimize
PV system designs or to provide production guarantees; investors look to verify
the right return on investment; operators need performance expectations to
compare to measured performance.
ll sectors of the maturing solar industry demand

accurate production estimates, which require
a clear understanding of how the estimates
are produced and an ability to interpret the
results. In this article we provide an overview of productionmodeling theory and review available production-modeling
tools. We compare the tools performance to each other and
to real systems, and provide a summary of the key uses of
production modeling in PV projects.
At the most basic level, production modeling comes
down to two questions:
1. How much sunlight falls on an array?
2. How much power can a system produce with

that sunlight?
Answering these questions requires location-specific
parameters, such as shading and weather data; educated
assumptions about system derating due to soiling, module mismatch, system availability; and complex algorithms to model
available radiation as well as module and inverter performance.
HOW MUCH SUN?
A PV systems geographical location, surroundings and configuration determine the amount of sunlight that falls on the
modules. Where a system is located geographically determines
how much sunlight is available; the surroundings dictate the
amount of available sunlight that is blocked before reaching the
array; and the array configuration determines how efficient the
system is at exposing the modules to sunlight.
Meteorological data. The first factor in determining how much
sunlight falls on an array is meteorological data that accurately
represent the weather at a systems location. Meteorological
data typically include solar radiation (global horizontal, direct
beam and horizontal diffuse), temperature, cloud cover, wind
speed and direction, along with other meteorological elements.
The data are based on ground or satellite measurements and in
some instances are modeled rather than measured.
30
S o l a r Pr o | April/May 2010
Typically a large amount of analysis is involved in taking

raw data and producing a data set suitable for use. Meteorological data are typically measured by government agencies
and utilized by a variety of organizations that make the data
available in formats suitable for use in production-modeling
tools. These organizations include the National Renewable
Energy Laboratory (NREL) and NASA, which provide the
information free of charge, and also organizations such as
Meteonorm and 3Tier, which provide the data for a fee.
The most common sources of data for US solar projects
are the Typical Meteorological Year (TMY) files published by
NREL and based on analysis of the National Solar Radiation
Data Base (NSRDB). TMY data comprise sets of hourly values
of solar radiation and meteorological elements representing a
single year. Individual months in the data record are examined,
and the most typical are selected and concatenated to form a
year of data. Due to variations in weather patterns, these data
are better indicators of long-term performance rather than
performance for a given month or year. According to the online
document Cautions for Interpreting the Results that NREL
publishes along with its PVWatts tool (see Resources), these
data may vary as much as 10% on an annual basis and 30%
on a monthly basis.
The first TMY data set was published in 1978 for 248 locations throughout the US. The data set was updated in 1994 from
the 19611990 NSRDB to create a set of TMY files, called TMY2,
for 237 US locations. A subsequent 2007 update utilized an
expanded NSRDB from 19992005 to create TMY3, which covers 1,020 locations across the US. TMY3 data are categorized
into three classes that reflect the certainty and completeness of
the data, with Class I being the most certain, Class II less certain and Class III being incomplete data. TMY, TMY2 and TMY3
present changes in reference time, format, data content and
units from set to set. The data sets are incompatible with each
other, but conversion tools are available. The TMY2 and TMY3
data sets are either utilized by or can be imported into all of the
major PV performance-modeling tools used in the US.
By Tarn Yates and Bradley Hibberd
w ww. f s im a ge s . c o m
deling
Radiation models. Typical weather data include three solar

radiation values representing radiation incident on a horizontal surface: direct beam, horizontal diffuse and global horizontal radiation. Direct beam radiation is light that travels in a
straight line from the sun, whereas diffuse radiation is light that
is scattered by the atmosphere or by clouds. In theory, global
horizontal radiation is the sum of the direct beam and the horizontal diffuse radiation. However, this is not always the case
due to measurement inaccuracies and modeling techniques.
Meteorological data indicate how much radiation falls
on a horizontal surface, but how much falls on an array?
While occasionally installed flat, PV systems usually have a

tilt and an azimuth or employ single- or dual-axis trackers. A

mathematical model is needed to translate horizontal radiation values into plane-of-array (POA) irradiance. The accuracy of a radiation model is affected by the weather at the
system location and by the quality of the weather data.
Numerous models are used to make this translation,
including the Perez et al., Reindel, Hay and Davies, and Isotropic Sky models. The Perez et al. model is the most complex.
A test performed in Albuquerque, New Mexico, by Sandia
showed that Perez et al. model predictions are the closest
to measured data. This is documented in the Sandia article
Comparison of PV System Performance-Model Predictions
solarprofessional.com | S o l a r P r o 31
Production Modeling
PV production models are really

quite simple. Making an accurate
model is straightforward. The difficult part is getting the right input
assumptions that drive the model
the most critical of these, of course,
being insolation.
Joe Song,
director of engineering,
SunEdison
with Measured PV System Performance (see Resources).

In general, radiation models treat
the direct beam component the same
way. Using the latitude and longitude of
the system location as well as the time
of day, it is possible to calculate the suns
position in the sky. Once this is known,
the translation of direct beam radiation
to POA radiation is a relatively simple
geometric calculation.
Where the models differ is in the treatment of diffuse
radiation. The Isotropic Sky model assumes diffuse radiation is emitted equally from every portion of the sky. More
advanced models take into account the fact that diffuse
radiation is more intense at the horizon and in the circumsolar region, the area directly surrounding the sun. They
may also consider variations in intensity based on the altitude angle of a section of sky, the clearness and brightness
of the sky, and the air mass. Refer to Solar Radiation and
Daylight Models for a history and review of radiation models (see Resources).
An additional component of radiation is the radiation
reflected by the ground or by the roof or surfaces associated
with the ground or roof. The reflected radiation is a function
of the albedo of the surface, a term that describes the reflective qualities of a surface. The amount of reflected radiation
is also a function of the angle of the array; an array at zero
degrees will receive no reflected radiation. The amount of
radiation received from reflection will increase with increasing tilt angle. Albedo varies with the surface and can change
throughout the year with weather conditions such as snow.
Modeling programs give you a variety of methods to account
Direct
from sun
m o ll y oh a l lo r an . c om
Diffuse
from sky
Fixed horizontal
collecting surface
Global horizontal radiation According to NRELs Glossary
of Solar Radiation Resource Terms, while total solar radiation
is the sum of direct, diffuse and ground-reflected radiation,
the amount of radiation reflected off of the ground is usually
insignificant. As a result, global horizontal radiation is generally referred to as the sum of direct and diffuse radiation.
32
for this. For example, both PVsyst

and PV*SOL allow you to define
monthly values for the albedo,
whereas the Solar Advisor Model
(SAM) changes the albedo if the
weather data indicate snow.
Shading. Simply translating
horizontal radiation into POA
radiation does not tell the whole
story. Depending on the PV system
location and configuration, large
distant objects, close obstructions and the system itself may
block some of the available sunlight. The complexity of the
performance-modeling tool dictates whether these types of
shading are treated separately or grouped together. In the latter case, shading is accounted for by a single derate factor.
Using a single derate factor for shading assumes that the
system experiences the same losses due to shade for every
hour of the year. In addition, most production-modeling tools
assume that the effects of shade are linear. That is, if 10% of
the array is shaded, then you lose 10% of the expected energy
production. This is not an accurate model, because shading
just one cell in a module can disproportionately impact the
whole module, the string or even the entire array.
Accurately defining shading is very difficult. It is not
possible to simply go out to a proposed project location,
look around and determine a shading derate factor. This
is where tools like the Solmetric SunEye and Solar Pathfinder are useful, because these tools quantify shading factors that can be used in many of the production-modeling
tools. Both Solmetric and Solar Pathfinder have their own
production software that is designed to interact with data
collected using their shade survey tools. (For more information on this topic, see Solar Site Evaluation: Tools and
Techniques to Quantify & Optimize Production, December/January 2009, SolarPro magazine.)
Soiling. An additional factor that decreases the available
sunlight is soiling caused by the accumulation of particulates, such as dust, snow, pollutants and bird droppings. The
power lost due to soiling is affected by the tilt of the array,
the quantity and seasonal variability of rain and snowfall, the
systems cleaning schedule and any site-specific conditions,
such as the proximity to a major roadway or a commercial
operation that creates dust. Most tools allow you to enter
an annual soiling derate factor only. This is not sufficient if
the value of power is determined by the period of time in
which the power is produced. For example, estimates for the
production losses due to soiling in California can be around
1% in winter and at least as high as 10% in late summer for a
system that is not washeda significant loss during a prime
production period that an annual soiling factor would not
accurately take into account. c o n t i n u e d o n p a g e 3 4
0/17/34/62
PMS 1805c
0/91/100/23
PMS 1805c
0/7/39/17
PMS 4525c
11/0/66/2
PMS 585c
65/0/100/42
PMS 364c
0/28/100/6
PMS 124c
11/1/0/64
PMS 431c
30/4/0/31
PMS 5425c
82/76/100/30
PMS 440c
0/46/100/33
PMS 7512c
3/0/100/58
PMS 385c
100/57/0/40
PMS 295c
Production Modeling
HOW MUCH POWER?
PV Performance Models
Several models have been created
to predict the power output of a
solar cell, module or array. Both
complex and simple models exist.
Here we describe some of the
more relevant models.
Sandia performance model. In
2004, Sandia National Laboratories published Photovoltaic
Array Performance Model, which Quantifying shade Solmetrics recently released PV Designer software tool allows you to
outlines the Sandia array perfor- drag icons representing data collected by its SunEye tool onto a visual representation of a
mance model (see Resources). This roof surface.
is one of the more robust production models. The Sandia performance model is based on a real-world losses, such as current leaks and resistance
series of empirically derived formulas that define five points between the metallic contacts and the semiconductor.
on the IV curve of a PV cell. These five points can be used to
Using circuit theory, you can define equations that
produce an approximation of the actual curve. The model describe the current and voltage characteristics of the
requires approximately 30 coefficients that are measured on a equivalent circuit. Unknown variables can be determined by
two-axis tracker at the Sandia National Labs in Albuquerque, evaluating the equations at conditions such as those specified on the manufacturers spec sheet for open-circuit voltNew Mexico.
The coefficients used in the Sandia model take into con- age and short-circuit current. The single-diode performance
sideration module construction and racking technique, solar model is the basis of both the model used in PVsyst and the
spectral influences, angle of incidence effects and the irra- CEC model that is an option in SAM.
diance dependence of electrical characteristics such as the
PVFORM model. The performance model that PVWatts uses
temperature coefficients of power, voltage and current. Tests is a simplified version of a model developed at Sandia called
documented in Comparison of Photovoltaic Module Perfor- PVFORM. This model uses the POA irradiance, ambient temmance Measurements show that the model can predict power perature and wind speed to calculate the operating temperaoutput to within 1% of measured power (see Resources).
ture of a solar cell. It then calculates the power output of the
The Sandia performance model is an option in both Solar system by adjusting the STC capacity rating of the array based
Advisor Model (SAM) and PV-DesignPro. One of the chal- on the POA irradiance and the cell temperature. As implelenges associated with this model is that the modules must mented in PVWatts, this model assumes that the temperature
undergo testing at the Sandia labs to be included. Unfor- coefficient of power for a PV module is -0.5%/C. This is a reatunately, this means that the Sandia database of modules sonable approximation for crystalline silicon modules that
often does not include recently released modules. This issue have temperature coefficients in the -0.55 to -0.40%/C range.
should soon be alleviated, as Sandia entered an agreement However, it is not appropriate for other technologies, such as
to have commercially available modules tested by TV thin film, that typically have temperature coefficients in the
Rheinland Photovoltaic Testing Laboratory at its facilities in -0.26 to -0.20%/C range.
Tempe, Arizona.
Single-diode performance model. The single-diode model DC Derate Factors
assumes that the behavior of a PV cell can be simulated by The major factors that determine the amount of dc power
an equivalent circuit consisting of a current source, a diode produced for a given level of illumination are the efficiency
and two or three resistors, as shown in Figure 1. The cur- of the technology, the temperature of the module cells and
rent source and diode represent the ideal behavior of a solar the technologys response to changes in temperature. Other
cell, and the series and shunt resistors are used to model factors that should be considered for accurate production
34
C o u r t e s y s o l m e tr i c . c o m
The second step in production

modeling is determining how
effective a PV system is at converting the sunlight incident on
an array into usable power.
I
ID
IShunt
modeling are the accuracy of
warranties for crystalline modRSeries
the nameplate rating of the
ules, such as the 85% power
module, losses due to module
guarantee after 25 years offered
IL
RShunt
V
mismatch, voltage drop across
with Suntechs Reliathon modthe diodes and connections in
ule, indicate that manufacturers
the modules, the resistance of
expect the value to be less. Addithe dc wiring, module degradationally, Comparison of Degradation, the inverters accuracy at Figure 1 This diagram shows the solar cell equivalent tion Rates of Individual Modules
tracking the maximum power circuit used in the single-diode performance model.
Held at Maximum Power (see
point of the array and the angle The current from the current source, IL, is directly
Resources) suggests that 0.5% per
of incidence of the sunlight.
year is a better rule of thumb for
proportional to the intensity of the available light and
Once the theoretical power the corresponding photoelectric effect.
crystalline modules, but notes
output of the array has been
that it should be higher than 1%
calculated, a series of derate
for many thin-film modules.
factors must be applied to arrive at the actual power that
will be delivered to the inverter. Following are some of the AC Derate Factors
major factors.
Unfortunately, the conversion of dc power delivered to the
Module nameplate rating. Module manufacturers assign a inverter into ac power at the point of interconnection is not a
range of accuracy to the nameplate rating of their modules, lossless process. The inverter is the major factor in this stage,
such as +/-5%. This means that a module rated at 200 W may but it is also important to consider losses due to wiring, transhave a power output of only 190 W. Unless the tolerance is formers and system downtime.
-0%, many modules do not have an STC rating as high as that
AC wiring losses. As with dc wiring, the losses due to resisspecified. A conservative value to use for this factor is one that tance in ac wiring vary with the amount of current. In the case
assumes that all of the modules have a rating at the low end of of ac current, loss factor calculations typically assume full
the tolerance.
power output from the inverter. This occurs for only a portion
DC wiring losses. Most integrators have standards for of the inverters operating time.
acceptable voltage drop that provide a good starting point
Transformer losses. When a transformer that is not
for determining this number. It is common for a wiring loss included as part of the inverter is required, it is necessary
factor to be calculated using the current and voltage at the to account for its losses. While many transformers are
maximum power point at STC conditions, as specified on the more than 98% efficient, it is worth verifying the transmanufacturers data sheet. Less rigorous tools take this single formers efficiency.
factor and apply it over all operating conditions. This practice
System downtime. Every PV system experiences downtime
neglects the fact that the current and voltage are rarely equal at some point. This can be due to the failure of an inverter or
to the values specified on the spec sheet. More advanced pro- a short in a single string. The severity and duration of the
grams (such as PVsyst, PV*SOL and PV-DesignPro) ask you to downtime can be mitigated by diligent maintenance, monispecify the size of conductors and length of the wire run, or toring and rapid response.
specifically ask for the losses at STC. They then calculate the
wiring losses at other operating conditions.
Inverter Performance Models
Module mismatch. This derate factor accounts for the fact According to the authors of Sandias Performance Model for
that the current and voltage characteristics of every module Grid-Connected Photovoltaic Inverters (see Resources), Freare not identical. Although the MPPT in the inverter keeps quently in modeling PV system energy production, inverter
the array at its maximum power point, each individual mod- efficiency is assumed to be a constant value, which is the same
ule does not operate at its maximum power point. A loss of as assuming that inverter efficiency is linear over its operat2% is a typical estimate for module mismatch. (Note that ing range, which is clearly not the case. In reality, the inverter
this factor is not relevant when using microinverters.)
efficiency depends on both the loading of the inverter and on
MPPT efficiency. According to Performance Model for the input voltage of the array. This is illustrated in Figure 2
Grid-Connected Photovoltaic Inverters (see Resources), most (p. 36), which shows a typical inverter efficiency graph availgrid-tied PV inverters are between 98% and 100% efficient at able through the CEC. A similar graph is available for every
capturing the maximum available power from a PV array.
inverter that is approved for incentives in California. An accuDegradation. If you are modeling future production, the rate inverter model should account for any power shaving that
degradation of power over time must be considered. A stan- may occur due to overloading or inverter shutdown due to the
dard value for module degradation is 1% per year. Recent dc voltage being out of range. The power consumption of the
Production Modeling
100
Efficiency (%)
Sandia performance model for grid-connected

PV inverters. The Sandia inverter model is similar
95
660 Vdc
740 Vdc
960 Vdc
90
to the Sandia module model in that it is based

85
on empirically derived equations. It considers
the ac power output of an inverter to be a func80
tion of the dc input power and voltage. Several
75
coefficients are used to define this relationship.
It is possible to approximate a version of the
70
inverter model with parameters usually avail0
100
10
20
30
40
50
60
70
80
90
able on a manufacturers spec sheet. Field and
Percent of rated output power
laboratory testing enable more refined versions
of the inverter model. A benefit of this model Figure 2 This graph is typical of the performance test results available for
is that it is compatible with the parameters all CEC-eligible inverters, showing, in this case, how the efficiency of an AE
recorded as part of the CEC testing process, and Solaron 333 is a function of inverter loading and dc input voltage.
therefore the associated database is kept upto-date. A Sandia study showed this model to be accurate to PVWatts
within 0.2% when compared to measured results. The Sandia PVWatts was developed by NREL and has long been the
inverter model is available in the system production-modeling default production-modeling tool of the US PV industry. Its
tool SAM.
strength lies in its simplicity. You can make a reasonable estiOther inverter models. The single-point efficiency model is mate of a systems production by selecting the location from
utilized in PVWatts and is also an option in SAM. This model a US map, entering the system size in dc watts and specispecifies a conversion efficiency that is used for all operating fying the array tilt and azimuth. You can also select singleconditions. In PVsyst and PV*SOL, inverters are defined by or dual-axis tracking options. By default the program uses
the manufacturers spec sheet values, such as the maximum a single conservative derate factor. This value is based on
power rating, the MPPT voltage range, the threshold power assumptions for variables such as the inverter efficiency, ac
and the inverters efficiency at various levels of loading. and dc wiring loses, and soiling. You can easily revise these
These programs use the efficiency inputs to define a curve assumptions to recalculate the derate factor.
PVWatts provides estimates of the monthly and annual
that is used in simulations. Although not a perfect correlavalues
for the ac energy production and average solar radiation, input values for defining inverter efficiency curves can
tion
per
day, plus a rough calculation of the value of the energy
be pulled from the online results of the CEC inverter tests at
produced
based on local energy rates. These values are often
Go Solar California, as illustrated in Figure 2.
reasonable estimates, but PVWatts lacks the level of control
and specificity of results that can be found in other tools.
Version 1. PVWatts v. 1 presents a simple map of the US
from which to choose the state where the project is located.
You then chose the TMY2 data location that is closest to the
While it is beyond the scope of this article to compare all of project site (in some instances the closest data location may
the available production-modeling tools, we review the major not be in the same state). A feature specific to v. 1 is that it
software packages currently utilized by researchers, integra- outputs an 8,760 reportan hour-by-hour report of energy
tors and project developers in North America: PVWatts, Solar production for the entire yearin text format.
Advisor Model, PV-DesignPro, PV*SOL and PVsyst.
Version 2. PVWatts v. 2 provides a map of the US that
These production-modeling tools, along with five oth- is divided into 40-by-40 km grid areas. The program then
ers, are surveyed in the companion table, 2010 Production- combines data from the closest TMY2 data location with
Modeling Tools, on pages 4043. This table does not include monthly weather data that are specific to the grid area
estimators used by various incentive or rebate programs and that you select. This more accurately reflects local weather
tools that are primarily intended to generate sales quotes conditions and accounts for distances from the TMY2
and proposals. Some of the entries in this table are adopted data locations. The v. 2 map is searchable by zip code or
from a table developed by Geoffrey Klise and Joshua Stein by latitude and longitude. A beta version of a new PVWatts
for their article Models Used to Assess the Performance of v. 2 map viewer was recently released. This new interface
Photovoltaic Systems (see Resources.)
allows you to quickly see the annual and monthly irradiance
PHOTOVOLTAIC
PRODUCTION-MODELING TOOLS
36
C o ur te sy g o so l a rc a l i fo r ni a . c o m
inverter under standby and operating conditions is also a factor in total power production.
specific to each grid cell. It is also easier to navigate and

more attractive.
HomeProHalfpageHorizontal010210.indd 1
14.0
13.0
0
10
20
30
40
50
Tilt
Parametric analysis The results from the parametric analysis optimization tool in SAM show that the tilt resulting in the
minimum levelized cost of energy (LCOE) is 32.5 with an
LCOE of 19.15 /kWh. This graph assumes a cash purchase,
using the default system cost and financial information provided in SAM. The system modeled consists of 1,190 Sharp
ND-216U1F modules with a due south azimuth connected to
a SMA Sunny Central 250U inverter in San Francisco, CA.
1/2/10 17:33:04
C o u r te sy nre l . g o v / a na l y si s/ sa m
15.0
LCOE
Solar Advisor Model (SAM)

SAM was produced by NREL in conjunction with Sandia
through the US Department of Energys Solar Energy Technologies Program. It is a step up from PVWatts in the level of
control available. SAM provides a wide range of options for
estimating PV module production, including the Sandia PV
array performance model, the CEC performance model and
the PVWatts performance model. The Sandia inverter performance model is used to simulate inverter performance.
You can select modules and inverters from databases so that
the specific characteristics of the system components can
be used in the simulations. In cases where components are
not in the databases, simple efficiency models can represent
their performance. SAM uses two composite derate factors,
pre-inverter and post-inverter, to account for system losses.
A 12-month-by-24-hour matrix is used to define the percent
of shading for every hour of every month of the year.
In addition to its production-modeling capabilities, SAM
puts an emphasis on analyzing the financials involved in PV
project development. The analysis focuses on the US market
16.0
and includes tax credits, depreciation, and capacityand production-based incentives. Detailed cash flow
models are available for residential, commercial
and utility-scale projects that can be used to calculate parameters such as the levelized cost of energy
(LCOE). SAM provides a method for entering utility
rate schedules, including time of use (TOU) schedules,
to accurately represent the varying value of electricity.
SAM contains a suite of analysis tools that includes
parametric, optimization, sensitivity and statistical
tools. These tools give you insight into how changes in
system variables (including tilt, azimuth, system capacity or component cost) impact output metrics such as
annual production or LCOE. The parametric and optimization tools run numerous iterations of the production simulation, stepping through a range of values that
you can define for one or more system variables. The
optimization tool maximizes or minimizes a specified
output metric, whereas the parametric tool provides a
broader view of the relationship between system variables and output metrics.
Two interesting new features were added to the
program with the release of the latest version in
October 2009. A scripting language called SAMUL has
been developed for SAM that is similar to the VBA
language available in Microsoft Excel. This allows you
to control many of the program functions through
code, and it facilitates the automation of repetitive
tasks. In addition, the program now generates source
code in Excel/VBA, C and MATLAB formats so that
the core simulation engine can be accessed separately from the user interface.
Co u r t e s y m a u i s o l a r s o f t w a re . c o m
Production Modeling
PV-DesignPro scatter plots These plots, with the hour of the day
and the solar irradiance on the horizontal plane and the array power
in dc watts on the vertical axis, show the difference in production for
a horizontal single-axis (north-south) tracker and a fixed system with
a tilt of 37 and an azimuth of 0 (true south) in San Francisco, CA.
Each figure shows 8,760 data points, one for every hour of the year.
(System specifications: 1,376 Mitsubishi PV-UD185MF5 modules;
one Xantrex PV225 inverter.)
PV-DesignPro
PV-DesignPro was developed by Maui Solar Energy
Software. The program is similar to SAM in that you
define system configuration and derate factors. PVDesignPro utilizes the Sandia PV array performance
model and provides module and inverter databases from
which to choose system components. The program accounts
for shading by means of a horizon profile that you define by
specifying the azimuth and altitude angle as well as the opacity of the obstruction. You also have the ability to define the
size and length of wire runs, as well as the efficiency of the
inverters MPPT. All other system losses are accounted for in
overall current and voltage derate factors.
One of PV-DesignPros strengths is the wealth of information that it supplies. At every step in the process the program attempts to provide as much insight as possible into
the variables that affect energy production. Once you select
a system location, for example, the program produces charts
showing detailed irradiance, temperature and wind data for
38
every day of the year. When defining system capacity, graphs

show typical IV curves and the max power of the array at cell
temperatures from 25C to 50C. Once you have run a simulation, you can create scatter plots containing data on system variables for every hour of the year. These scatter plots
can be used to visualize and learn about system behavior or
to inform design decisions.
PV-DesignPro also performs parametric analyses and
produces graphs that illustrate how changes in system variables influence production and financial parameters. This
function can help you minimize or maximize important
variables such as kWh production or the cost of a utility bill.
The software also includes tools to produce detailed load
and TOU profiles. These can be used to c o n t i n u e d o n p a g e 4 4
Production Modeling
2010 Production Modeling Tools 1

Basics
Software
Program
HOMER
Developer
Cost
free
Polysun
HOMER ENERGY,
originally
developed by
NREL
Vela Solaris
Light $159
Pro $489
PV Designer
Solmetric
$400/yr
PV-DesignPro
WebBased or
Application
application
Weather Data Source
Irradiance Model
user provides hourly average global solar radiation on the horizontal

surface (kW/m2), monthly average global solar radiation on the
horizontal surface (kWh/m2/day), or monthly average clearness index
Hay and Davies model
application
Meteotest
unknown
application
various weather sources including TMY2 and TMY3 data; outside the
US, the same weather sources as Energy Plus
Perez et al. model
Maui Solar
$259
Energy Software
with Sandia
F-Chart Software $400
with University
of Wisconsin
Valentin
$698 2
Software
application
TMY2, TMY3 , Meteonorm, Global Solar Irradiation Database
Perez et al. model (default), HDKR

model (option)
application
TMY2, TMY3, weather data can be added
Isotropic Sky model
application
MeteoSyn, Meteonorm, SWERA, PVGIS, NASA SSE
Hay and Davies model
PVsyst
University of
Geneva
application
NREL
PVWatts v. 2
NREL
free
Web
RetScreen
Natural
Resources
Canada
NREL
free
application
TMY2, TMY3, Meteonorm, ISM-EMPA, Helioclim-1 and -3, NASA-SSE,

WRDC, PVGIS-ESRA and RETScreen; user can import custom data in
a CSV file
in the USTMY2 data; 239 options outside the USTMY data from
the Solar and Wind Energy Resource Assessment Programme, the
International Weather for Energy Calculations (V1.1), and the Canadian
Weather for Energy Calculations
combination of TMY2 data with monthly weather data from Real-Time
Nephanalysis (RTNEPH) database (cloud cover), Canadian Center
for Remote Sensing (albedo), National Climatic Data Center (daily
maximum dry bulb temperatures) and RDI/FT Energy (1999 residential
electric rates)
combination of weather data collected from 4,720 sites from 20
different sources with data from 19611990 & NASA-SSE
Hay and Davies model (default),

Perez et al. model (option)
PVWatts v. 1
1st license
$984,
additional $197
free
free
application
TMY2, TMY3, EPW, Meteronorm
Perez et al. model (default);

Isotropic Sky Model, Hay and
Davies model, Reindl model
(options); total and beam (default),
beam and diffuse (option)
PV F-Chart
PV*SOL
Solar Advisor
Model (SAM)
Web
Notes:
1
Some entries in this table adopted from Klise and Stein (2009). 2 Does not include expert version to be released in 2010.
3
Shading derate is from SunEye readings. Inverter efficiency derate is from an equipment database.
4
User enters array operating temperature, reference efficiency, temperature coefficient and array area.
40
Perez et al. model
Perez et al. model
Isotropic Sky model
Modeling
Production-Estimating Model:
Module
Production-Estimating Model:
Inverter
Simulation
Frequency
linear irradiance model with

temperature correction
single efficiency derate factor
hourly
manual
input
manual
input
derate factors not categorized, all losses except for

single percentage for inverter efficiency are covered by
miscellaneous losses
empirical model of module

performance, dependent on three
MPPT power ratings at different
irradiance values and the module
temperature coefficient
proprietary model based on nominal
power and operating temperature
unknown
hourly
manual
input
manual
input
soiling, degradation, mismatch, wiring
single-weighted efficiency
derate factor
hourly
manual
input
manual
input
Sandia model
Sandia model
hourly
manual
input
manual
input
PV module nameplate dc rating, inverter and transformer,

mismatch, diodes and connections, dc wiring, ac wiring,
soiling, system availability, shading, sun tracking, age 3
wiring, MPPT efficiency, array current derate factor, array
voltage derate factor
function of efficiency and

temperature
power tracking and power

conversion efficiency factors
hourly
manual
input
manual
input
inverter conversion efficiency and power tracking efficiency
modeled using V and irradiance at

STC, module efficiency curve and
an incident angle modifier; linear or
dynamic temperature model options
Shockleys one-diode model for
crystalline silicon; modified onediode model for thin film
simplified PVFORM
inverter profile and efficiency

curve generated from measured
data
hourly
manual
input
manual
input
mismatch, diodes, module quality, soiling, wiring, deviation

from standard spectrum, module height above ground
inverter profile and efficiency

curve generated from measured
data
hourly
manual
input
manual
input
hourly
manual
input
manual
input
field thermal loss, standard NOCT factor, Ohmic losses,

module quality, mismatch, soiling (annual or monthly), IAM
losses
soiling, system availability, shading, sun tracking, age
simplified PVFORM
monthly
manual
input
manual
input

soiling, system availability, shading, sun tracking, age
Evans average efficiency model
monthly
manual
input
manual
input
inverter efficiency, miscellaneous losses
Sandia model, CEC model, PVWatts

model
single efficiency derate factor,

hourly
Sandia Model for grid-connected
inverters
manual
input
manual
input
mismatch, diodes and connections, dc wiring, soiling, sun

tracking, ac wiring, transformer
Tilt
Orientation
Derate Factors
Production Modeling
2010 Production Modeling Tools

Modeling
Software
Program
HOMER
Technologies
Tracking
Output Data
single axis (horizontal, daily adjustment), single

axis (horizontal, weekly adjustment), single axis
(horizontal monthly adjustment), single axis
(horizontal, continuous adjustment), single axis
(vertical, continuous adjustment), dual axis
cSi, aSi, CdTe, single axis, dual axis
CIS, CIGS, HIT,
c-Si, Ribbon
(EFG)
cSi, aSi, CdTe, n/a
CIS
not considered independently, could be

incorporated into single derate factor
hourly ac production data
horizon profile may be defined or imported
unknown
sub-module level shading, computed based

on distance-weighted interpolation of
readings taken from Solmetric SunEye
cSi, aSi, CdTe, single axis (horizontal axis EW), single axis (horizontal horizon profile user-defined
CIS, CPV,
axis NS), single axis (vertical axis), single axis (NS
mj-CPV
axis parallel to Earths axis), dual axis
not
flat-plate array, single-axis tracking (adjustable tilt/
not considered, could be incorporated into
technology
azimuth), dual-axis tracking, concentrating parabolic other derate factors
specific 4
collector
hourly ac energy production; daily and

monthly ac energy production displayed
graphically on screen
hourly data available for meteorological
data, PV array behavior (cell temp, module
efficiency), energy production and more
monthly average hourly values of ac
energy
horizon profile user-defined or imported

from shade survey tool, 3D modeling
environment in Expert version
horizon profile can be user-defined or
imported from a shade survey tool, 3D
modeling environment, based on array
configuration
hourly energy production in one-week

segments
PVWatts v. 1
cSi, aSi, CdTe, single axis (vertical), dual axis

CIS, HIT,
c-Si, Ribbon
cSi, HIT, CdTe, single axis (horizontal axis EW), single axis (vertical
aSi, CIS, c-Si axis), single axis (tilted axis), dual axis, dual axis
(frame NS), dual axis (frame EW), tracking sun
shields; ability to define parameters such as collector
width, shade spacing and rotation limits
cSi
single axis, dual axis
single derate factor
hourly ac energy production
PVWatts v. 2
cSi
single derate factor
n/a
RetScreen
cSi, aSi,
single axis, dual axis, azimuth
CdTe, CIS,
spherical-Si
cSi, aSi, CdTe, single axis (tilted NS axis), dual axis
CIS, CPV, HIT
n/a
n/a
12-month by 24-hour shade profile can be

imported

data, PV array behavior (cell temp, wiring
losses, etc.), energy production
Polysun
PV Designer
PV-DesignPro
PV F-Chart
PV*SOL
PVsyst
Solar Advisor
Model (SAM)
not
technology
specific4
Shading
single axis, dual axis
Notes:
4
User enters array operating temperature, reference efficiency, temperature coefficient and array area.
42
n/a = not available

data, PV array behavior (cell temp, wiring
losses, etc.), energy production
Details
Financial Analyses
cash-flow analysis considering
energy costs, operating costs and
calculation of LCOE
Component Database
Ability to Export
Data to Excel
exported as a
text file
Optimization
Module
Inverter
Update Method and Frequency
sensitivity
analysis and
optimization
capability
n/a
n/a
n/a
User Support &

Documentation
user manual provided with
software
financial analysis including O&M

costs, incentives, projected
electricity costs, inflation and
interest rates
n/a
yes
n/a
yes
yes
automatically checks for updates

software
yes
n/a
yes
yes
component data complied from PVXchange

database, updated approximately monthly

software
basic cash-flow analysis
yes
parametric
analysis
yes
yes
online help file, training

videos
parametric
analysis
n/a
n/a
updates supplied periodically on the Maui

Solar Software site; you can add modules
and inverters
n/a
tilt, inter-row
yes
spacing, inverter
loading
tilt, orientation, yes
inter-row
spacing, inverter
loading
yes
lifecycle cost calculations including can be copied and

electricity purchased from
pasted into Excel
utility, electricity sold to utility,
O&M costs, rebates, tax credits,
depreciation; cash-flow analysis
economic efficiency and cash-flow yes
analysis

software
considers energy costs, feed-in

tariffs and system financing
yes
basic calculation of energy value
n/a
n/a
n/a
basic calculation of energy value
8,760 report is
output as text that
can be pasted into
an Excel file
n/a
updates to the database are supplied by

manufacturers; the program can be set to
check for updates at start up
updated approximately once a year, usually
with the release of a software update; you
can define additional components or import
individual component files received from
other sources
n/a
n/a
n/a
n/a
n/a
detailed cash-flow analysis,

sensitivity and risk analysis
program is Excel
based
n/a
yes
n/a
manufacturer must contact RetScreen
detailed cash-flow analysis for

residential, commercial and utility
scale projects; focused on the US
market; sensitivity and statistical
analysis tools
yes
numerous
yes
production
and financial
optimization
tools, parametric
analysis
yes
CEC module model (NREL maintains a

extensive user manual,
library of CEC-approved modules), SAM can detailed help file, online user
sync with the most recent library, additional group, email support
modules can be added by contacting NREL;
library of inverter coefficients is updated
regularly as the CEC inverter database is
updated
yes
limited help file available with

program; training available
detailed help file available
with program, FAQ on Web
site, no user manual
online documentation and

support available
limited help file provided

available with program,
additional online documentation and support available
online manual, detailed help
file, online training courses
PV-DesignPro parametric analysis This chart was created

using the default load profile available in PV-DesignPro and
the PG&E A-6 rate schedule that is preloaded in the program.
The lowest electric bill for a customer in San Francisco, CA, is
achieved at a module tilt of 30 and an azimuth of 10. (System
specifications: 1,376 Mitsubishi PV-UD185MF5 modules; one
Xantrex PV225 inverter.)
compare the financial benefits that may result from switching rate schedules when installing a PV system.
PV*SOL
PV*SOL is produced by Valentin Software, based in Germany.
The program is widely used in the European market, and Valentin has begun efforts to increase market share in the US. These
efforts include a 2010 release of an Americanized version of
both PV*SOL and its most advanced tool, PV*SOL Expert, that
use American numbering conventions and a North American
product database. PV*SOL contains an extensive database of
modules and inverters that is frequently updated. The program
can be set to automatically check for updates to the database
on startup. You can account for shading by creating or importing a horizon profile. Derate factors, such as mismatch, soiling,
dc voltage drop, module tolerance, and losses across diodes
and connections, are all considered.
At the start of each session you are given the option to use
a Quick Design tool. After you select a specific type of module,
the number of modules that are to be installed and an inverter
brand, the program calculates all of the possible stringing
combinations. The options are ranked based on how efficient
they are at using inverter capacity. This is useful when trying
to determine the best way to use numerous string inverters
on a project.
PV*SOL stands out in its ability to model multiple arrays
and multiple inverters in the same simulation, something not
possible with most tools. Each array can be specified independently of the others, including module type, array tilt and
azimuth, and single or multiple inverters. Derate factors and
horizon profiles can also be specified independently for each
44
C o u r t e s y v a l e nti n- so ftw a re . c o m
Co ur te sy m a ui so l a r so ftw a re . c o m
Production Modeling
PV*SOL shading simulation This PV*SOL screen capture

is color-coded to indicate the amount of shading across the
roof. The numbers on the modules indicate the shading loss
for each. A US version of PV*SOL will be available in 2010.
array. On complex projects with multiple buildings, this can significantly reduce the simulation time.
PV*SOL Expert contains a 3D shade modeling environment in which a building can be defined that includes typical
features such as gables and chimneys. Other objects that may
shade an array, such as trees and additional structures, can
be added to the model. You can then run a simulation that
color-codes the roof according to the amount of shade an area
receives. This simulation also lets you arrange modules on the
roof and see the shading loss for each one, as shown in the
screen capture above.
Although many of the advanced tools available in both
versions of PV*SOL are geared toward the simulation of
roof-mounted systems, the program also contains options
for vertical single-axis tracking as well as dual-axis tracking.
The program does not have an option for horizontal singleaxis tracking.
PVsyst
PVsyst, developed at the University of Geneva, Switzerland,
is currently the hot name in production modeling. It is
the primary tool used by independent engineers who are
brought in to verify production numbers for investors. The
program contains a large database of modules and inverters for component selection. PVsyst considers many of
the system losses as the other modeling tools do. Where it
stands out is its treatment of shading and soiling.
You have the ability to enter a different soiling factor
for each month in PVsyst, which more accurately reflects
real-world conditions. The program can quickly model the
effects of inter-row shading through c o n t i n u e d o n p a g e 4 6
Production Modeling
an option called unlimited sheds that calcuIn order to provide an

PVsyst provides more conservalates when the system experiences inter-row
understanding
of the relative
tive results and is more powerful
shading based on the array parameters and
performance of each tool
at covering complex issues such
on the location and orientation of the array.
in different scenarios, we
as shading.
PVsyst also provides you with a 3D CADcompare the performanceManfred Bchler,
like environment in which a more complex
modeling tools production
chief technical officer,
model of a PV system and the nearby surestimates for crystalline
Phoenix Solar
roundings can be created. Once an array is
silicon PV modules on a
defined, it can be broken into strings, and
fixed-tilt array, a single-axis
the effect that shading has on a string can be specified.
tracking array and a dual-axis tracking array, as well as thinPVsyst provides numerous array configuration options. film modules on a fixed-tilt array.
To simulate tracking, you can define the important charTo perform the simulations in each modeling tool across
acteristics such as single or dual axis, maximum and mini- the three mounting systems and the two module technologies,
mum tilts, the spacing between rows or arrays, and whether we input specifications for four generic systems, as follows:
or not the tracker employs backtracking. (Backtracking is
a tracking strategy controlled by a microprocessor that Crystalline Systems
adjusts the array tilt to constantly avoid inter-array shadModules: Sharp ND-216U2 (216 W STC, 187.3 W PTC)
ing, especially early and late in the day.) PVsyst can simulInverter: Xantrex GT250 (250 kW, 96% CEC efficiency)
taneously model systems that comprise more than one size
Array: 1,400 modules (302.4 kW STC), 100 strings of
or type of inverter, as well as arrays with two different tilts
14 modules each
and azimuths connected to a single inverter.
Installation #1: Fixed-tilt ground mount, 0 azimuth (true
What makes PVsyst such a valuable tool is not that it has
south), 30 tilt
a more accurate model for PV or solar cell production than
Installation #2: Single-axis tracking (north-south),
the other production-modeling systems available, but rather
0 azimuth (true south)
its unique ability to control and accurately define many of
Installation #3: Dual-axis tracking
the other factors that are involved in production modeling.
The report that PVsyst produces, and in particular the dia- Thin-Film System
gram showing system losses, is especially valuable. A new
Module: First Solar FS255 (55 W STC, 51.8 W PTC)
version of the program, PVsyst 5.0, was released in June 2009
Inverter: Xantrex GT250 (250 kW, 96% CEC efficiency)
and updates to the program are released regularly on the
Array: 5,028 modules (276.5 kW STC), 838 strings of
PVsyst Web site (see Resources).
6 modules each
Installation: Fixed-tilt ground mount, 0 azimuth (true
south), 30 tilt c o n t i n u e d o n p a g e 4 8
We use the production-modeling tools just discussed to simulate the annual energy yield for different system designs. In
this section we compare the tools production estimates for
theoretical systems of different technologies and perform
two case studies to compare the modeling tools production
estimates to measured production. These tools are evaluated in the following model-to-model comparisons:
PVWatts, v. 1
PVWatts, v. 2
PVsyst v. 4.37
SAM, Sandia PV performance model and Sandia

inverter performance model
SAM, CEC PV performance model and Sandia

inverter performance model
PV*SOL 3.0, release 7
PV-DesignPro, v. 6.0
46
Co u r t es y pv s ys t . co m
COMPARISON OF
PV PRODUCTION MODELS
PVsyst 3D model The near shading scene function in PVsyst

is used to calculate the impact of obstructions like adjacent
trees or structures on system performance. In this case, the
effects of shading are modeled on a vertical east-west singleaxis tracking system.
Production Modeling
The systems are sized by starting with a chosen inverter,

dividing the ac power rating by the CEC-rated efficiency, then
dividing by the modules PTC rating. The resulting number of
modules is rounded up to a whole number of strings.
Modeling-Tool Parameters
We use the default derate parameters for each modeling
toolwith the exception of SAM, for which we match
the derate factors to those from PVWatts for consistency. Table
1 lists the derate parameters used in the various modeling tools.
Each PV system is located in San Francisco, California.
NREL TMY2 data for that location are used in the modeling.
For the purposes of modeling with PVWatts v. 2, the 94124
zip code is used to identify the 40-by-40 km grid.
Each tools default POA radiation model is used. This
means that simulations performed with PVWatts v. 1 and
v. 2, SAM and PV-DesignPro use the Perez et al. model; PVsyst
and PV*SOL use the Hay and Davies model.
To maintain consistency between tools when modeling
tracking, we did not use PVsysts capability to model the backtracking or shade avoidance. In addition, the horizontal
single-axis tracking design was not modeled in PV*SOL, as
that tool can model only a vertical single-axis tracking design.
Results of Model-to-Model Comparisons
The results of the modeling comparisons are presented in
terms of specific yield in Graph 1. Specific yield is the production in kWh with respect to the STC system size in kW. In other
words, it is energy divided by nameplate power. This allows
for a more direct comparison between different technologies.
In reviewing the results presented in Graph 1 and the
source data, we make the following observations about the
estimates that each of the tools generated:
For any single scenario, the discrepancy between the
maximum and minimum production estimate ranged
from 9% to 14%; the average difference was 11.5%.
The largest discrepancy between production estimates

was 14% for the thin-film scenario. This reflects the
greater level of uncertainty associated with modeling the
performance of thin-film modules.
With the exception of the thin-film scenario, PV*SOL and
PVWatts (v. 1 and v. 2) consistently produce estimates
that fall between those for SAM and PV-DesignPro at the
high end and PVsyst at the low end.
In the thin-film scenario, the relatively lower estimates
for PVWatts v. 1 and v. 2 are expected due to the inability
of the tool to accurately model thin-film performance.
What is unexpected is that the PVsyst estimate is similar
to those from PVWatts v. 1 and v. 2.
The estimates of the two SAM models were consistently
the largest or most aggressive estimates. Using the CEC PV
performance model, SAM generally estimated a 1% higher
annual production than it did when using the Sandia PV
array performance model. The small percentage suggests
that the difference in module performance models is
small, in the context of a full-system simulation.
PV-DesignPro consistently estimates between 1.5% and
2% below the SAM models, but still significantly higher
than most other tools estimates. By default, PV-DesignPro
considers MPPT efficiency and dc wire loss only. We expect
that its production estimates would be lower if consistent
derate factors were applied.
PVsyst consistently produced the smallest or most conservative production estimates. Comparing the PVsyst
loss diagram that the software generates with the simple
derate factors for other modeling tools leads us to believe
that this result is largely due to the module performance
model within PVsyst. Differences in module and inverter
characteristics within the tools databases may also contribute to this result.
PVWatts v. 1 estimates an average of 2% more annual production than v. 2. We believe the difference is attributable
Derate Factors Model-to-Model Comparisons

PV module nameplate
Inverter & transformer
Mismatch
Diodes & connections
dc wire loss
ac wire loss
Soiling
Shading
Sun tracking
MPPT efficiency
48
PVWATTS
v. 1 & v. 2
0.95
0.96
0.98
0.995
0.98
0.99
1
1
1
-
SAM (CEC &

Sandia models)
MOD
0.98
0.995
0.98
0.99
1
MOD
1
-
PVsyst
PV*SOL
PV-DesignPro
0.97
MOD
0.98
MOD
MOD
1
1
1
MOD
-
1
MOD
0.98
0.995
MOD
1
1
1
1
-
1
MOD
1
1
.99
1
1
1
0.95
Table 1 Derate factors for

each program are translated to a decimal value
for comparison, matching
the convention used in
PVWatts. MOD denotes
that the parameter is modeled within the tool, rather
than reduced to a single
derate factor.
2,200
1,800
1,600
PV-DesignPro
PV*SOL
SAM - CEC
PVSyst
1,000
SAM - Sandia
1,200
PVWatts v. 2
1,400
PVWatts v. 1
C o u r t e s y b o r re g o s o l a r. c o m
Specific yield (kWh/kWp)
2,000
30 tilt
Crystalline silicon
Single-axis tracking
Crystalline silicon
Dual-axis tracking
Crystalline silicon
30 fixed tilt
CdTE thin film
Graph 1 This graph shows the annual specific yield estimated by the different PV production models for the four comparison
PV systems. Absent data in the single-axis tracking example is due to the fact that PV*SOL does not model vertical (northsouth) tracking.
to the modification of weather data in PVWatts v. 2 to

improve geographic resolution; as such, other sites may
produce dissimilar results.
CASE STUDIES: COMPARING MODELING

TOOL OUTPUT TO PRODUCTION DATA
To compare predicted performance with the measured performance of actual systems, we perform two case studies of
PV systems in operation. Case Study #1 is a fixed-tilt hybrid
monocrystalline /amorphous silicon installation on a rooftop in Escondido, California. Case Study #2 is a fixed-tilt carport installation with amorphous silicon thin-film modules
in Santee, California. Both projects have monitoring equipment that includes measurement of insolation; as such,
both the energy produced by the systems and the insolation
available to the systems can be compared to simulations.
For the case studies, we reduced the number of tools
used. This is due to the similarity in results observed in the
comparisons between two pairs of PVWatts and SAM models. For PVWatts, only v. 2 was used in the case studies. For
the two SAM models, we used the Sandia PV array performance model for Case Study #1 and the CEC performance
model for Case Study #2; this is due to the availability of
modules in the respective databases.

Modeling Parameters
Weather data. The meteorological data for all simulations are
NREL TMY2 data for San Diego, California, with the exception of the PVWatts v. 2 simulation, which uses modified
data based on the zip code for each system.
Shading. Each modeling tool addressed inter-row shading as follows:
In PVsyst, by utilizing the unlimited sheds modeling
technique;
in SAM by using the 12-by-24 shading matrix;
in PVWatts by entering the shading loss resulting from
the PVsyst simulation; and
in PV*SOL and PV-DesignPro by creating a horizon
profile.
No additional shading is considered, because the arrays
are largely shade-free.
Soiling. This is modeled in PVsyst at 1.5% per month, accumulating from month to month when the average rainfall in
that month was not significant. When rainfall was significant
or the system was cleaned, the soiling factor was reduced to
1.5% for that month. Case Study #1 was not cleaned and the
resulting annual soiling loss was 4%. Case Study #2 was cleaned
at the end of June, and the resulting annual soiling loss was
3.1%. These annual soiling losses are used in all modeling tools.
Other. Except as noted below, all other derate factors are
as per Table 1:
Production Modeling
is close to the measured production, with the exception of

the PV*SOL modeling tool. The combination of the modeled
insolation being lower than measured, but modeled production approximately matching what was measured, indicates
that the modeling tools will significantly overestimate system
production if an average or typical weather year were to occur.
Our interpretation is that the system is underperforming with
respect to the modeling tools predictions. This underperformance is consistent with reports from the project site indicating that significant soiling is reducing production.
Graph 2 shows that the monthly production estimates
and measured production values are within the same range
and follow the same trend over the course of the year, with
some exceptions. The most significant exception is the drop
in measured production in June. When reviewing the insolation data, we observe an equivalent drop. Therefore the
system is performing as expected. (This drop in June is also
observed in Case Study #2.)
With the exception of June, the modeling tools appear to
have produced estimates in reasonable c o n t i n u e d o n p a g e 5 2
Energy (kWh)
Case Study #1
The first case study is a 78.4 kW roof-mounted array in
Escondido, California, consisting of Sanyo HIP-200BA3
hybrid monocrystalline/amorphous silicon modules that
are tilted at 10 and oriented directly south (0). The array
is wired with seven modules per source circuit, and the
resulting 56 source circuits are connected to a PV Powered
PVP75KW-480 inverter. The system has been in operation
for just over 18 months with no significant downtime since
being commissioned. The site is relatively new construction and is located in an area
where further construction is
occurring. As a result, soiling is
Measured
SAM (Sandia)
PVWatts
expected to have a significant
13,000
impact on the systems performance. In addition, there is a
12,000
local wastewater ordinance
11,000
restricting the owners ability
to clean the system. Therefore,
10,000
it has not been cleaned since it
PVSyst
PV-DesignPro
was commissioned.
9,000
Results. The modeling results
8,000
for Case Study #1 are presented
in Table 2. They show that mea7,000
sured insolation is approxiPV*SOL
mately 10% greater than mod6,000
eled. This is consistent across
the different tools, indicating
Jan.
Feb.
Mar.
Apr.
May
June
July
Aug.
Sept.
Oct.
Nov.
Dec.
that they perform comparably
in modeling weather data. The Graph 2 This graph shows the monthly energy production in kWh for the measured and
estimated production, however, modeled system in Case Study #1.
Case Study #1: Measured-to-Modeled Comparison

Insolation (kWh/m2/year)
Delta to measured (%)
Production (kWh)
Measured
2,178.6
0.0%
123,058
0.0%
PVsyst
1,977.3
-9.2%
119,816
-2.6%
SAM (Sandia)
1,981.2
-9.1%
127,107
3.3%
PVWatts
2,004.8
-8.0%
119,986
-2.5%
PV*SOL
1,911.8
-12.2%
114,736
-6.8%
PV-DesignPro
1,984.6
-8.9%
118,502
-3.7%
Table 2 This table presents the measured and estimated annual insolation and production values for Case Study #1 as well as
the percent difference of measured-to-modeled values.
50
C ou r t e s y b o rreg os o l a r. c o m
In PV*SOL a module tolerance of -3% is specified.

In PV-DesignPro MPPT efficiency is modeled as 98%; an
array voltage derate factor of 0.975 is used to account for
module mismatch and losses in diodes and connections;
wiring losses are set at 3%.
As these systems are both in their first 1218 months
of operation, no module degradation is considered. System
availability is also not considered, because each system had
no significant downtime.
Production Modeling
The wide variation is an indicator that modeling the performance of thin-film modules is more complex and presently
less accurate than modeling performance for crystalline silicon modules.
PVWatts is limited in its ability to model modules other
than crystalline silicon. Given that amorphous silicon
modules are used in this case study, we account for this
limitation in PVWatts by applying a correction factor to
the STC system size specified in the PVWatts model. The
correction factor is determined by comparing the PTC to
STC ratio for the Kaneka G-SA60 module to that for a reference crystalline module, in this instance the Sharp ND216U2. The PTC to STC ratio is 10% higher for the Kaneka
module; as a result, the system size modeled in PVWatts
is increased by 10%. The results shown in Table 3 indicate
that the adjusted PVWatts v. 2 results are similar to those
for the other tools. This approach is similar to the one used
by the Los Angeles Department of Water and Power in its
incentive program. While this appears to produce reasonable
results, more effective tools are available for modeling thinfilm module performance.
Energy (kWh)
Case Study #2
The second case study is a 481.5 kW carport-mounted array
in Santee, California, consisting of Kaneka G-SA60 singlejunction amorphous silicon thin-film modules, tilted at 5
and oriented 27 west of true south. The array is wired with
five modules per source circuit, and the
resulting 1,605 circuits are connected to
Measured
two Xantrex GT250-480 inverters. The
95,000
carport is actually an RV parking shelter
PVSyst
SAM (CEC)
PVWatts
and has a roof deck immediately below
85,000
the modules, which reduces airflow and
PV*SOL
increases module temperature. The sys75,000
tem has been in operation for just over
12 months with no significant downtime since being commissioned.
65,000
Results. The modeling results for
PV-DesignPro
Case Study #2 are presented in Table 3.
55,000
They show that measured insolation is
approximately 5% lower than modeled.
45,000
This is consistent across the different
tools, indicating that they model weather
data comparably. The estimated proJan.
Feb.
Mar.
Apr.
May
June
July
Aug.
Sept.
Oct.
Nov.
Dec.
duction, however, varies widely, ranging
from 3% below the measured value for Graph 3 This graphs shows the monthly energy production in kWh for the meaSAM to 15.2% below for PV-DesignPro. sured and modeled system in Case Study #2.
Case Study #2: Measured-to-Modeled Comparison

Insolation (kWh/m2/year)
Production (kWh)
Measured
2,037.6
0.0%
849,136
0.0%
PVsyst
1,944.1
-4.6%
779,192
-8.2%
SAM (CEC)
1,922.7
-5.6%
823,635
-3.0%
PVWatts
1,956.4
-4.0%
777,359
-8.5%
PV*SOL
1,855.7
-8.9%
759,531
-10.6%
PV-DesignPro
1,918.3
-5.9%
719,869
-15.2%
Table 3 This table presents the measured and estimated annual insolation and production values for Case Study #2 as well as
the percent difference of measured to modeled values.
52
C o u rt es y b or reg o s o l a r. c o m
agreement with the measured data. However, when you

examine Graph 2 closely, you can see thatwith the exception of Junethe measured data either exceed or are equal
to the estimated data from January to July. It is reasonable
to suppose that if insolation in June had not been relatively
low, the production that month would also have exceeded
the predictions. From August through October, however,
the measured data fall below nearly all of the modeled estimates. Only one modeled data pointthat for PVsyst in Septemberis lower than the measured data. This indicates the
impact of soiling on production through the dry summer
season in San Diego County. The PVsyst capability to model
soiling on a monthly basis captures the behavior. The estimated production values in November and December are
similar to the measured values.
Graph 3 shows that the monthly estimates for production

and the measured production follow the same broad trend, in
terms of an increase in production during the summer. As in
Case Study #1, the one instance where measured and modeled
production do not track one another is the drop in measured
production in June. Again, the insolation data reveal a similar
reduction, and thus the behavior is as expected.
While generally predicting near the average of the other
modeling tools, PVsyst has the highest production estimate
in July. This is due to PVsysts ability to model month-bymonth soiling factors. The soiling factor was reduced from
6% for June to 1.5% for July when scheduled cleaning was carried out, and the resulting production increase is reflected in
the production graph. Other tools also show a similar trend,
but this is simply in proportion to the increased insolation
available in July.
THE VALUE OF
PRODUCTION MODELING
Production modeling impacts many aspects of PV project

development. During the sales cycle, performance estimates
are necessary for determining project capacity and lining up
financing. These estimates are also used during the design
and engineering phase to make informed design decisions
that optimize PV system performance. During operations,
production modeling is used to evaluate system performance to ensure appropriate production. Production modeling also has a key role in the evaluation of new products
and technologies.
System sizing. Production estimates of varying complexity are essential in determining the appropriate size system
to build. In simple situations where customers are trying to
offset a portion of their annual energy bill, a back-of-theenvelope production estimate may suffice. However, if customers are trying to zero out their electric bill or if TOU rate
schedules are in play, the method used to estimate production
needs to be more precise, more sophisticated. You can have
more confidence in design decisions by modeling with tools
that use location-specific weather data and produce hourly
estimates of production.
Financials. Revenue from energy production is a major force,
if not the driving force in PV project development. In an environment where the majority of PV projects, particularly larger
projects, are not purchased outright but financed through
complex deals, the value of each kWh generated cannot be
understated. Incentives based on kWh rather than kWsuch
as the California Solar Initiative Performance Based Incentive
program or one of many solar renewable energy credit programscan double or triple the simple value of a kWh, exceeding $0.30/kWh.

Production Modeling
Given the potential value of each kWh, system

Once a technology
production has a huge impact on the revenue a
choice has been made,
project generates. If production is significantly
modeling tools allow you
under- or overestimated, the effects can be serito optimize the array
ous on the project at hand, on future deals and
layout. A general rule of
on the industry as a whole.
thumb holds that the
Underestimated production can cause
optimal configuration to
any number of development issues, perhaps
maximize annual producmisrepresenting project viability or resulting
tion is a tilt angle equal to
in an oversized system. Underestimated prothe sites latitude with a
duction may prevent a project from being developed that due south azimuth. While this rule would be true for a singlemight otherwise have been attractive. Or it could push a plane array under ideal circumstances, inter-row shading and
customer toward a deal with a developer whose production local weather variations can skew the optimum configuration.
estimate is higher. If an oversized system results, the excess Modeling tools can be used both to find the optimal configuelectricity generated may have to be given away to the util- rations and to look at what effect a nonoptimal configuration
ity without compensation.
would have. For fixed-tilt systems, modeling tools can be used
Overestimated production may result in changes to the to determine the effects of inter-row shading. They also help to
financial structure of the project. This is true when the com- determine the balance between the increased capacity allowed
missioned system cannot meet the performance require- by smaller shade setback distances and the decreased producments established through production modeling. Production tion. For tracking systems, modeling tools can help you make
guarantees that are based upon an overestimated production decisions about the spacing of arrays or whether backtracking
model can lead to financial penalties for the party guarantee- is a valuable option. The 3D shade simulations can be used to
ing the system performance. An underperforming asset may place arrays in areas where they are least impacted by shading
not have the market value that an owner had planned on from trees or roof obstructions.
when committing to the project terms.
Performance-modeling tools also allow you to make
Whether used by investors examining revenue streams, informed decisions about inverter sizing. For example, if a
integrators looking to guarantee that revenue, or end cus- building can accommodate an array rated at 500 kW STC,
tomers looking to offset their utility bills, accurate energy should you use a 500 kW inverter or a 350 kW inverter?
production estimates are crucial to all parties in the suc- Using a modeling tool that accounts for power loss due to
cessful deployment of a solar energy project. Given this clipping allows you to compare the value of the lost power
importance, investors rarely evaluate production estimates themselves. Instead, independent engineering firms with extensive
production-modeling experience are generally
relied upon. Typically, the independent engineering firm also verifies system performance
following commissioning.
System design. Production-modeling tools
play an essential role in maximizing the production or financial return of a PV system. The
first step is making a decision about what technology to deploy based on a given location or
a set of financial considerations. Different climates and locations affect the output of various technologies, such as crystalline silicon
versus thin-film PV or single- versus dual-axis
trackers. The times of the day and seasons of
the year when these technologies produce Inverter clipping This PV-DesignPro scatter plot has one data point for
power also vary. A technology that has the each hour of the year. It illustrates how much power clipping results
best financial return in one location or under a from overloading a Xantrex PV225 inverter with a 384.8 kW array (2,080
given rate schedule may not be the best choice Mitsubishi PV-UD185MF5 modules) for a system in San Francisco, CA,
with a 25 tilt and a 0 (true south) azimuth.
in other circumstances.
54
Co u rt es y m au i s o l ar s o f t wa re .c om
Currently, all the models lack the

seriousness that can be provided
only by having skin in the game.
Once there is a tool out there that
people put money behind, the
entire solar industry will get far
more serious and real.
Fred Unger, president,
Heartwood Group
The Dollars Are in the Details
Da t a c o u r t e s y n re l . g o v / a n a l y s i s / s a m
he following production-modeling examples, which seek

to correlate annual production to system tilt and azimuth,
show the importance of using modeling tools that account for
detailed system variables.
Example 1: SAM. An optimization run using SAM for a
250 kW system in San Francisco, California, at a latitude of
37.6, shows that annual production is maximized with a tilt
Example 2: PVsyst. Production numbers run in PVsyst,

which provides an inter-row shading option, show that for systems with an 8 azimuth and inter-row spacing that keeps the
array shade free between 9am and 3pm on December 21,
a tilt angle of 25 actually produces slightly more annual
power than one tilted at 32.5. This is illustrated in Graph 5,
which shows the monthly kWh production for 25 and 32.5
tilt angles, as modeled by PVsyst. Tilting the array at 25 has
additional benefits: Production is weighted toward the summer
months when power is generally more valuable; the system
covers a smaller area; and less racking material is required.
In this case, using the data from SAM would appear to result
in a less productive, more expensive system. You could run
additional comparisons to optimize the system for total production, TOU weighted production or other system metrics. {
45
25 tilt
40
35
30
25
over the life of the system when using the 350 kW inverter to
the increased upfront cost of installing the 500 kW inverter.
You can run the same type of analysis to make the decision
between a single inverter or multiple inverters for arrays
with different orientations.
Operations. Production-modeling tools can also be used to
evaluate a PV systems long-term performance. Accurate production modeling establishes a relationship between the irradiance available to the system and the electricity produced by
the system. This ratio is applied to the measured irradiance
and used to determine the expected production. This result
can be compared to the measured production to determine
whether the system is performing as expected. This can be
done in real time, typically using Web-based analysis tools for

Dec.
Nov.
Oct.
Sept.
Aug.
July
June
May
Apr.
Mar.
Feb.
Jan.
20
of 32.5 and an azimuth of 8, where true south is 0 and

positive values indicate an azimuth that is west of south. See
Graph 4 for a representation of this result. The SAM optimization assumes no shade. However, most large systems are
composed of numerous rows spaced at a calculated distance,
and are often designed to have inter-row shading before 9am
and after 3pm on December 21. Unfortunately, SAM does not
provide an easy method for defining inter-row shading.
Graph 5 This graph was produced using monthly energy

production numbers generated by PVsyst. It indicates that for a
system with an 8 azimuth in San Francisco, CA, a 25 tilt generates more energy than a 32.5 tilt, especially in the summer.
viewing the data from the system, or retrospectively over a

given time, typically monthly or annually. Accurate modeling
of all of the system parameters is critical to the effectiveness
of this technique, as are accurate measurements of the irradiance and production values.
CONCLUSIONS
Based on our evaluations, the radiation model components

of the evaluated tools perform consistently, predicting similar POA irradiance from the same weather data. In terms
of production estimates, SAM is the most aggressive modeling tool and PVsyst the most conservative. There is an
average of 9% difference between their estimates.
Data courtesy pvsyst.com
Graph 4 This contour graph was created by SAM and

shows the relationship of energy production to tilt and azimuth for a modeled PV system in San Francisco, CA.
Energy (MWh)
32.5 tilt
Production Modeling
New technologies and applications

create new challenges for modelers. There
is a continuing need for development
and validation of models for diverse
technologies, applications and climates
to ensure model accuracy and to quantify
uncertainty.
Chris Cameron,
project lead for systems modeling,
Sandia National Laboratories
Given the importance of accurate energy production estimates,

the sophistication and capabilities
of modeling tools must continue to
evolve along with the solar industry. At this stage, an ideal tool might
combine the following features: the
Sandia PV array performance model;
a component database updated as
frequently, or more often, than the
CEC database; PVsysts control over
system and location variables; and
SAMs ability to perform financial, parametric and statistical
analyses. Throw in the ability to define 3D layouts in a CADlike environmentas in PVsystand to load shade readings
taken in the fieldas with Solmetrics PV Designer software
and its SunEye tooland you would have it all.
In the end, production-modeling tools are only as good
as the person who uses them. The choice of derate factors
can easily shift a production estimate by 5% or more. That
said, for accurate simulations, it is important

to have a tool that gives
you as much control as
possible over the factors
that affect production.
Currently, PVsyst is the
tool that stands out, due
to its ability to account
for shading from a variety of sources and to
vary soiling definitions
over the course of the year as well as its flexibility to model a
large number of different configurations.
The authors wish to thank Geoffrey T. Klise and Christopher P. Cameron of Sandia National Laboratories for their
expert input during preparation of this article as well as for
sharing a prepublication draft of the report Models Used to
Assess the Performance of PV Systems.
g C O N TAC T
Tarn Yates / Borrego Solar / Berkeley, CA /
[email protected] / borregosolar.com
Bradley Hibberd / Borrego Solar / Berkeley, CA /
[email protected] / borregosolar.com
R eso u rces
Production-modeling software
HOMER / 720.565.4046 / homerenergy.com
Publications
Cautions for Interpreting the Results, NREL /
rredc.nrel.gov/solar/codes_algs/PVWATTS/
interp.html
Comparison of Degradation Rates of Individual
Modules Held at Maximum Power,
C.R. Osterwald, J. Adelstein, J.A. del Cueto,
B. Kroposki, D. Trudell, T. Moriarty, NREL Report:
PR-520-39844, May 2006 / nrel.gov/publications
Polysun / 415.671.6292 / velasolaris.com

PV Designer / 707.823.4600 / solmetric.com
PV-DesignPro / mauisolarsoftware.com
PV F-Chart / 608.255.0842 / fchart.com
PV*SOL / 888.786.9455 / valentin-software.com
PVsyst / +41.22.379.0650 / pvsyst.com
PVWatts v. 1 / 303.275.3000 / rredc.nrel.gov/
solar/codes_algs/PVWATTS/version1/
PVWatts v. 2 / 303.275.3000 / rredc.nrel.gov/
solar/codes_algs/PVWATTS/version2/
RetScreen / 613.995.0947 / retscreen.net
Solar Advisor Model (SAM) / 303.275.3000 /
nrel.gov/analysis/sam/
56
Comparison of Photovoltaic Module

Performance Measurements, A.H. Fanney,
M.W. Davis, B.P. Dougherty, D.L. King, W.E.
Boyson, and J.A. Kratochvil, Journal of Solar
Energy Engineering, Vol. 128, May, 2006
Comparison of PV System PerformanceModel Predictions with Measured PV System
Performance, Christopher P. Cameron,
William E. Boyson, Daniel M. Riley, Sandia
National Laboratories, 33rd IEEE PV Specialist
Conference, San Diego, CA, 2008
Improvement and Validation of a Model for
Photovoltaic Array Performance, W. DeSoto, S.A.
Klein, W.A. Beckman, Solar Energy, Journal of
the International Solar Energy Society, Vol. 80,
Issue 1, January 2006 / elsevier.com
Modeling Daylight Availability and Irradiance

Components from Direct and Global Irradiance,
R. Perez, P. Ineichen, R. Seals, J. Michalsky,
R. Stewart, Solar Energy, Journal of the
International Solar Energy Society, Vol. 44,
Issue 5, January 1990 / elsevier.com
Performance Model for Grid-Connected
Photovoltaic Inverters, David L. King, Sigifredo
Gonzalez, Gary M. Galbraith and William E.
Boyson, Sandia Report: SAND2007-5036, 2007
Models Used to Assess the Performance of
Photovoltaic Systems, G.T. Klise and J.S. Stein,
Sandia Report: SAND2009-8258, 2009
Photovoltaic Array Performance Model, D.L.
King, W.E. Boyson, J.A. Kratochvill, Sandia
Report: SAND2004-3535, 2004
Solar Radiation and Daylight Models, Tariq
Muneer, 2004, Hardback, 392 pages, ISBN10: 0-7506-5974-2, $80.95 from Elsevier,
314.447.8010 / elsevier.com
Go Solar California / gosolarcalifornia.
ca.gov/equipment (Inverter performance test
summaries located here.)
Glossary of Solar Radiation Resource Terms /
rredc.nrel.gov/solar/glossary

Production Modelling Grid-Tied PV Systems

Uploaded by

Copyright:

Available Formats

Production Modelling Grid-Tied PV Systems

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Production Modelling Grid-Tied PV Systems

Uploaded by

Copyright:

Available Formats

Production Mo

for Grid-Tied PV Systems

ll sectors of the maturing solar industry demand

HOW MUCH SUN?

Typically a large amount of analysis is involved in taking

By Tarn Yates and Bradley Hibberd

Radiation models. Typical weather data include three solar

tilt and an azimuth or employ single- or dual-axis trackers. A

PV production models are really

with Measured PV System Performance (see Resources).

for this. For example, both PVsyst

HOW MUCH POWER?

The second step in production

Sandia performance model for grid-connected

to the Sandia module model in that it is based

specific to each grid cell. It is also easier to navigate and

Solar Advisor Model (SAM)

every day of the year. When defining system capacity, graphs

2010 Production Modeling Tools 1

Weather Data Source

user provides hourly average global solar radiation on the horizontal

Hay and Davies model

Perez et al. model

TMY2, TMY3 , Meteonorm, Global Solar Irradiation Database

Perez et al. model (default), HDKR

TMY2, TMY3, weather data can be added

Isotropic Sky model

MeteoSyn, Meteonorm, SWERA, PVGIS, NASA SSE

Hay and Davies model

TMY2, TMY3, Meteonorm, ISM-EMPA, Helioclim-1 and -3, NASA-SSE,

Hay and Davies model (default),

TMY2, TMY3, EPW, Meteronorm

Perez et al. model (default);

Perez et al. model

Perez et al. model

Isotropic Sky model

linear irradiance model with

single efficiency derate factor

derate factors not categorized, all losses except for

empirical model of module

soiling, degradation, mismatch, wiring

PV module nameplate dc rating, inverter and transformer,

function of efficiency and

power tracking and power

inverter conversion efficiency and power tracking efficiency

modeled using V and irradiance at

inverter profile and efficiency

mismatch, diodes, module quality, soiling, wiring, deviation

inverter profile and efficiency

field thermal loss, standard NOCT factor, Ohmic losses,

single efficiency derate factor

PV module nameplate dc rating, inverter and transformer,

Evans average efficiency model

single efficiency derate factor

inverter efficiency, miscellaneous losses

Sandia model, CEC model, PVWatts

single efficiency derate factor,

mismatch, diodes and connections, dc wiring, soiling, sun

2010 Production Modeling Tools

PVSOL shading simulation This PVSOL screen capture