Frequently Asked Questions#
General Questions#
What is pvlib?#
pvlib is a free and open-source python software library for modeling the electrical performance of solar photovoltaic (PV) systems. It provides implementations of scientific models for many topics relevant for PV modeling.
For additional details about the project, see Package Overview. For examples of using pvlib, see Example Gallery.
How does pvlib compare to other PV modeling tools like PVsyst or SAM?#
pvlib is similar to tools like PVsyst and SAM in that it can be used for “weather-to-power” modeling to model system energy production based on system configuration and a weather dataset. However, pvlib is also very different in that you use pvlib via python code instead of via a GUI, which makes pvlib ideal for automating tasks. pvlib is also more of a toolbox or a framework to use to build your own modeling process (although some pre-built workflows are available as well).
Usage Questions#
All I have is GHI, how do I get to POA?#
Going from GHI to plane of array (POA) irradiance is a two-step process. The first step is to use a decomposition model (also called a separation model) to estimate the DNI and DHI corresponding to your GHI. For a list of decomposition models available in pvlib, see DNI estimation models.
The second step is to transpose those estimated DNI and DHI components into
POA components. This is most easily done with the
pvlib.irradiance.get_total_irradiance() function.
Where can I get irradiance data for my simulation?#
pvlib has a module called iotools which has several functions for
retrieving irradiance data as well as reading standard file formats
such as EPW, TMY2, and TMY3. For free irradiance data, you may
consider NREL’s NSRDB which can be accessed using the
pvlib.iotools.get_psm3() function and is available for
North America. For Europe and Africa, you may consider looking into
CAMS (pvlib.iotools.get_cams()).
PVGIS (pvlib.iotools.get_pvgis_hourly()) is another option, which
provides irradiance from several different databases with near global coverage.
pvlib also has functions for accessing a plethora of ground-measured
irradiance datasets, including the BSRN, SURFRAD, SRML, and NREL’s MIDC.
Can I use PVsyst (PAN/OND) files with pvlib?#
Although pvlib includes a function to read PAN and OND files
(read_panond()), it is up to the user to determine
whether and how the imported parameter values can be used with pvlib’s models.
Easier use of these parameter files with the rest of pvlib may be added
in a future version. Until then, these Google Group threads
(one
and two)
may be useful for some users.
Why don’t my simulation results make sense?#
pvlib does not prevent you from using models improperly and generating
invalid results. It is on you as the user to understand the models you
are using and to supply appropriate, correctly-formatted data. One modeling error that beginners sometimes
make is improper time zone localization. Calculating solar
positions is often the first step of a modeling process
and this step relies on timestamps being localized to the correct time zone.
A telltale sign of improper time zones is a time shift between solar
position and the irradiance data (for example, solar_elevation
peaks at a different time from clear-sky ghi).
For more information on handling timezone correctly, see Time and time zones.
More generally, inspecting the simulation results visually is a good first step when investigating strange results. Matplotlib and pandas have very powerful plotting capabilities that are great for tracking down where things went wrong in a modeling process. Try plotting a few days of intermediate time series results in a single plot, looking for inconsistencies like nonzero irradiance when the sun is below the horizon. This will give you a clue of where to look for errors in your code.
I got a warning like RuntimeWarning: invalid value encountered in arccos, what does it mean?#
It is fairly common to use pvlib models in conditions where they are not applicable, for example attempting to calculate an IV curve at night. In such cases the model failure doesn’t really matter (nighttime values are irrelevant), but the numerical packages that pvlib is built on (e.g. numpy) emit warnings complaining about invalid value, divide by zero, etc. In these cases the warnings can often be ignored without issue.
However, that’s not always the case: sometimes these warnings are caused by an error in your code, for example by giving a function inappropriate inputs. So, these warnings don’t necessarily indicate a problem, but you shouldn’t get in the habit of immediately discounting them either.
I got an error like X has no attribute Y, what does it mean?#
If you see a function in the pvlib documentation that doesn’t seem to exist
in your pvlib installation, the documentation is likely for a different version
of pvlib. You can check your installed pvlib version by running
print(pvlib.__version__) in python. To switch documentation versions, use
the v: version switcher widget in the bottom right corner of this page.
You can also upgrade your installed pvlib to the latest compatible version
with pip install -U pvlib, but be sure to check the What’s New
page to see the differences between versions.
The CEC table doesn’t include my module or inverter, what should I do?#
The CEC tables for module and inverter parameters included in pvlib are periodically copied from SAM, so you can check the tables there for more up-to-date tables.
For modules, if even the SAM files don’t include the module you’re looking for
either, you can calculate CEC module model parameters from
datasheet information using pvlib.ivtools.sdm.fit_cec_sam().
Which should I use, the CEC or the Sandia PV Module database?#
The CEC PV module database contains parameters for significantly more modules, and is more up to date, than the Sandia PV module database. Therefore, the CEC PV module database is probably the more useful option in most cases. However, finding parameters for the specific module being used is usually more important than which database they came from.
Besides which modules each database includes, another consideration is the different modeling capabilities each parameter set provides. The CEC model produces a continuous IV curve while the Sandia model calculates only a few specific points of interest on the curve. For typical simulations where only the maximum power point is of interest, either model will suffice.
How do I model a system with multiple inverters?#
Currently, pvlib’s ModelChain and PVSystem only support simulating one inverter at a time. To calculate total power for multiple inverters, there are two options:
If the modules, mounting, stringing, and inverters are all identical for each
inverter, then you may simply simulate one inverter and multiply the
ModelChainResult.ac by the number of inverters to get the total system output.
If the inverters or their arrays are not all identical,
define one PVSystem and ModelChain per inverter and
run the simulation for each of them individually. From there you
can add up the inverter-level outputs to get the total system output.