Weather data#
Simulating the performance of a PV system requires irradiance and meteorological data
as the inputs to a PV system model. Weather datasets are available
from many sources and in many formats. The pvlib.iotools module
contains functions to easily retrieve and import such datasets in a standardized
form that is convenient to use with the rest of pvlib.
The primary focus of pvlib.iotools is time series solar resource
data like the irradiance datasets from PVGIS and the NSRDB, but it also provides
functionality for other types of data useful for certain aspects of PV modeling.
For example, precipitation data is available from get_acis_prism()
for soiling modeling and horizon profiles can be retrieved from
get_pvgis_horizon() for horizon shade modeling.
For a complete list of functions related to retrieving and importing weather data, see IO Tools, and for a detailed comparison of the weather datasets available through pvlib, see 1.
Types of weather data sources#
Weather data can be grouped into a few fundamental categories. Which type is most useful depends on the application. Here we provide a high-level overview of different types of weather data, and when you might want to use them.
Ground station measurements: From in-situ monitoring equipment. If properly maintained and quality-controlled, these are the highest quality source of weather information. However, the coverage depends on a weather station having been set up in advance for the location and times of interest. Some ground station networks like the BSRN and SURFRAD make their measurement data publicly available. An global overview of ground stations is available at https://SolarStations.org.
Data from public ground station measurement networks are useful if you want accurate, high-resolution data but have flexibility around the specific measurement location.
Satellite data: These sources process satellite imagery (typically from geostationary satellites) to identify and classify clouds, and combine this with solar irradiance models and aerosol data to produce irradiance estimates. The quality is generally much higher than NWP, but still not as good as a well-maintained weather station. They have high spatial and temporal resolution corresponding to the source satellite imagery, and are generally optimised to estimate solar irradiance for PV applications. Free sources such as PVGIS, NSRDB, and CAMS are available, and commerical sources such as SolarAnywhere, Solcast, and Solargis provide paid options though often have free trials.
Satellite data is useful when suitable ground measurements are not available for the location and/or times of interest.
Numerical Weather Prediction (NWP): These are mathematical simulations of weather systems. The data quality is much lower than that of measurements and satellite data, owing in part to coarser spatial and temporal resolution, as well as many models not being optimised for solar irradiance for PV applications. On the plus side, these models typically have worldwide coverage, with some regional models (e.g. HRRR) sacrifice global coverage for somewhat higher spatial and temporal resolution. Various forecast (e.g. GFS, ECMWF, ICON) and reanalysis sources (ERA5, MERRA2) exist.
NWP datasets are primarily useful for parts of the world not covered by satellite-based datasets (e.g. polar regions) or if extremely long time ranges are needed.
Usage#
With some exceptions, the pvlib.iotools functions
provide a uniform interface for accessing data across many formats.
Specifically, pvlib.iotools functions usually return two objects:
a pandas.DataFrame of the actual dataset, plus a metadata
dictionary. Most pvlib.iotools functions also have
a map_variables parameter to automatically translate
the column names used in the data file (which vary widely from dataset to dataset)
into standard pvlib names (see Variables and Symbols).
Typical usage looks something like this:
# get_pvgis_tmy returns two additional values besides df and metadata
df, _, _, metadata = pvlib.iotools.get_pvgis_tmy(latitude, longitude, map_variables=True)
This code will fetch a Typical Meteorological Year (TMY) dataset from PVGIS,
returning a pandas.DataFrame containing the hourly weather data
and a python dict with information about the dataset.
Most pvlib.iotools functions work with time series datasets.
In that case, the returned df DataFrame has a datetime index, localized
to the appropriate time zone where possible. Make sure to understand each
dataset’s timestamping convention (e.g. center versus end of interval), as
pvlib will use these timestamps for solar position calculations.
The content of the metadata dictionary varies for each function/dataset.
Data retrieval#
Several pvlib.iotools functions access the internet to fetch data from
online web APIs. For example, get_pvgis_hourly()
downloads data from PVGIS’s webservers and returns it as a python variable.
Functions that retrieve data from the internet are named get_, followed
by the name of the data source: get_bsrn(),
get_psm3(), get_pvgis_tmy(),
and so on.
For satellite/reanalysis datasets, the location is specified by latitude and longitude in decimal degrees:
latitude, longitude = 33.75, -84.39 # Atlanta, Georgia, United States
df, metadata = pvlib.iotools.get_psm3(latitude, longitude, map_variables=True, ...)
For ground station networks, the location identifier is the station ID:
df, metadata = pvlib.iotools.get_bsrn(station='cab', start='2020-01-01', end='2020-01-31', ...)
Some of these data providers require registration. In those cases, your access credentials must be passed as parameters to the function. See the individual function documentation pages for details.
Reading local files#
pvlib.iotools also provides functions for parsing data files
stored locally on your computer.
Functions that read and parse local data files are named read_, followed by
the name of the file format they parse: read_tmy3(),
read_epw(), and so on.
For example, here is how to read a file in the TMY3 file format:
df, metadata = pvlib.iotools.read_tmy3(r"C:\path\to\file.csv", map_variables=True)
References#
- 1
Jensen et al. “pvlib iotools—Open-source Python functions for seamless access to solar irradiance data”. Solar Energy, 2023. DOI: 10.1016/j.solener.2023.112092.