1.0.2 (July 22, 2021)


  • Added function to parse BSRN station-to-archive formatted data (modified version of a pvlib python function). (GH541, GH656)
  • Added functions to fetch data from the NASA Langley BSRN site. (GH541, GH656)
  • Added functions to parse data for the PNNL reference data site. (GH563, GH669)
  • Report timezone may now be set by passing the timezone keyword argument to ReportParameters. (GH317, GH681)
  • Report tables of observations, forecasts and reference forecasts now link users to the date range of the data used in the report. (GH679, GH688)


  • SRML and PVDAQ contained PV sites with positive temperature coefficients. These temperature coefficients were multiplied by -1. (GH625, GH634)
  • Added/removed requirements to make base, test, and fetch install options work as expected. (GH637, GH643)
  • Updated PyPI classifiers. (GH637, GH643)
  • Updated Installation page with instructions for PyPI, conda-forge, and docker. (GH635, GH649)
  • If total metrics category was not selected, the report now displays warnings instead of confusing blank tables. (GH629, GH652)
  • Removed the word “solar” from the report intro line for consistency with other forecast analyses such as net load and wind power. (GH624, GH650)
  • Removed the BSRN Granite Island MI site from reference database. This site was added without realizing 2020 and newer data is not readily available from the NASA web pages. (GH541, GH656)
  • Reports could not be created for event forecasts due to an issue with validating event forecast data. (GH660, GH661)
  • Changed GFS fetch directory for compatibility with March 22, 2021, 12Z model upgrade. (GH662, GH663)
  • Worked around data gaps during CLEARSKY GHI validation that cause uneven frequencies which lead to skipping validation entirely (GH673)
  • Fixed issue with processed observations failing to be plotted due to a missing forecast. Processed observations are now plotted for all points for which there is a forecast. (GH689)
  • Fixed probabilistic persistence ensemble forecasts to use the last 30 days of data and persistence_probabilistic_timeofday() rather than the last hour of data with persistence_probabilistic(). Hour ahead reference forecasts are replaced with day ahead reference forecasts. (GH639, GH645)
  • Clarified that automated generation of reference forecasts is currently limited to privileged accounts. (GH659, GH691)
  • Fixed reference forecast links in report tables (GH678, GH688)


  • Clarified expectations for pandas.DatetimeIndex.freq attribute throughout test suite. Improves compatibility with pandas >= 1.1. (GH641)

  • Test configuration additions (GH637, GH643):

    • Python 3.7, 3.8, 3.9 using the latest versions of the [all] dependencies
    • Python 3.7 and 3.8 using the minimum version of the [all] dependencies. These are typically the values from the requirements file that the docker image uses.
    • Python 3.7 using only the [test] dependencies