PVPMC_2025

Will Hobbs

This is a repository for my 2025 PVPMC poster, "Plant performance analysis with satellite resource data and public power data".

(view poster)

Code for pulling from gridstatus.io and matching with EIA was based on code by Drumil Joshi, Southern Power Company.

Why?

The goal of this work was to highlight a potentially valuable dataset from ERCOT: Power data are available for every solar plant in ERCOT at 15-min intervals back to 2012. I think someone should analyze it all and publish results.

This is just a small, informal preview of what could be done with that data.

Details

A copy of the poster is in the presentation folder. I suggest you check it out first.

The notebook Get_ERCOT_EIA_Data.ipynb gets plant metadata and timeseries power data from ERCOT and EIA. About 20 plants were initially selected semi-randomly with the following criteria:

• In service in 2022 or earlier
• Greater than 30 MW
• Then the first ~10 plants from each ERCOT region (alphabetical) with unique names were selected
• Plants where I had a known affiliation with the owner/operator were dropped (this is the non-random part)

I used https://github.com/gridstatus/gridstatusio to pull the timeseries data from https://www.gridstatus.io/.

Note: data pulled form gridstatus.io are not included in this repository. Users will need to get that data using their own API key.

The list of plants was then filtered to plants that were operating for all of 2021, reducing the number to 10 plants.

Here's a sample of power data:

The notebook Process_Specs_and_Data.ipynb converts specifications/metadata from ERCOT and EIA into parameters to model expected energy with the model_pv_power() function in pv_model.py from https://github.com/williamhobbs/pv-system-model and NREL NSRDB PSM4 weather data (pulled with the v0.12.1-alpha.1 release of https://github.com/pvlib/pvlib-python). It also includes some basic sample analyses.

Here's monthly performance index (PI) for the 10 plants, where 5 with subjectively large drops in PI are highlighted:

Notes:

Things I didn't do, but should have, include:

• run more plants
• use pvanalytics
  • clear sky filter, check tilt/az, try the new outlier check PR #223, ...
• use rdtools
• use solardatatools
• explore and explain the different data channels from ERCOT, like:
  • HSL = high sustainable limit, what the plant could procude when un-curtailed
  • Telemetered net output, what the plant actually produced
  • Basepoint, somehow a little different from telemetered net output??