search

openclimatefix / Open-Source-Quartz-Solar-Forecast

Open Source Solar Site Level Forecast
MIT License
43 stars 40 forks source link

Add visualization graphs to eval #102

Open JasonFengGit opened 3 months ago

JasonFengGit commented 3 months ago

Pull Request

Description

Add visualization graphs to eval. Fix the TODO in evaluation.py Including "Predicted VS Actual" graph for every 48hrs segments and "Error Distribution" for all tested instances.

How Has This Been Tested?

Checklist:

Remaining Issues

  1. My code in local env uses the pip package quartz_solar_forecast, and it seems my visualization code has to be updated in the package before I can use. Investigating how to solve this.
  2. How to update documentation for this?
  3. I realize a lot of timestamps in results_df does not have real generation_power values

Examples

error_distribution.png image

pred_vs_actual_9531_2021_07_28_22_00_00.png image

pred_vs_actual_9531_2021_09_16_07_00_00.png image

codecov[bot] commented 3 months ago

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 79.94%. Comparing base (05ae555) to head (3d9de9e). Report is 12 commits behind head on main.

Additional details and impacted files ```diff @@ Coverage Diff @@ ## main #102 +/- ## ======================================= Coverage 79.94% 79.94% ======================================= Files 12 12 Lines 354 354 ======================================= Hits 283 283 Misses 71 71 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

JasonFengGit commented 2 months ago

Hi @jacobbieker @peterdudfield, could you give any suggestions for the issues above? Thanks!

JasonFengGit commented 2 months ago

Hi @jacobbieker, is it a known issue that a lot of timestamps in results_df does not have real generation_power values? Thanks!

peterdudfield commented 2 months ago
  1. My code in local env uses the pip package quartz_solar_forecast, and it seems my visualization code has to be updated in the package before I can use. Investigating how to solve this.

You can use pip install -e . and hopefully you can then use it

  1. How to update documentation for this?

Is there something you can add on the readme about visualization of the evaluation

  1. I realize a lot of timestamps in results_df does not have real generation_power values

Hmm this is interesting, would you be able to write which values they are. It might be worth putting these in a different issue

JasonFengGit commented 2 months ago
  1. My code in local env uses the pip package quartz_solar_forecast, and it seems my visualization code has to be updated in the package before I can use. Investigating how to solve this.

You can use pip install -e . and hopefully you can then use it

  1. How to update documentation for this?

Is there something you can add on the readme about visualization of the evaluation

  1. I realize a lot of timestamps in results_df does not have real generation_power values

Hmm this is interesting, would you be able to write which values they are. It might be worth putting these in a different issue

Thanks! Let me address these issues. : )

JasonFengGit commented 2 months ago

Created an issue here https://github.com/openclimatefix/Open-Source-Quartz-Solar-Forecast/issues/110