Use kadi dynamic commanded states for starcheck thermal model

[jeanconn]

Use kadi dynamic commanded states for starcheck thermal model

Closes #329

Supersedes #319

[jeanconn]

@taldcroft I think this is all looking good, but what would you like to see as documentation of testing for both the PR and for LR?

As we've changed

• method of getting states for continuity interval
• handling of eclipses (from none to some handling)
• determination of initial state with temperature from telemetry

I'm not sure if each needs to be validated independently or if "acq and guide temps for a set of schedules match expectations" is enough.

[taldcroft]

Showing that "acq and guide temps for a set of schedules match expectations" is enough, where of course the set of schedules includes a couple with eclipses.

[jeanconn]

Regarding eclipses, I suppose I can use the proseco pkl files for straightforward independent determination of those expectations (I had just been going "oh that's different and probably right due to eclipse").

[taldcroft]

Or use one of the eclipse loads that failed starcheck but passed FOT thermal, and show now passing. I would approve that.

[jeanconn]

Regarding the states in general, I was happy to see the temperatures matched the MCC/pkl ones much better for the eclipse schedules, though I wasn't sure if we should add eclipse to the starcheck aca thermal model plot.

I also thought these were close to the mcc values but not quite as close as I would have guessed.

(and I still had an action to try to understand what is happening in the beginning of JUN1719A)

[taldcroft]

Re eclipse checking, looks great, I'll call that a pass!

[taldcroft]

Code looks good now, thanks! Just need to re-run the validation plots and standard regression.

[jeanconn]

Regarding standard regression, I ran with the "-3" run_start_time vs current master (not release) and get text diffs that look reasonable to me, with temperature differences that look up to 0.1 deg on first scroll-through:

regress_diffs.txt

[jeanconn]

In more review it looks like there are some larger temperature diffs. I notice some in JAN1215B which should have eclipses in it. I would prefer to do more rigorous eval of the diffs (and confirm that all larger diffs are due to eclipses) but I think basically we are OK and any reasonable diffs related to temperature are fine. We're mostly using the full regression set to just confirm that everything runs and there are no new weird errors.

[jeanconn]

For the validation plots, I ran these with these start times

./sandbox_starcheck -dir kadistates/jun1019a/ -out jun1019a_kadi -run_start_time 2019:155:14:00:00.000
./sandbox_starcheck -dir kadistates/jun1719a/ -out jun1719a_kadi -run_start_time 2019:162:22:00:00.000
./sandbox_starcheck -dir kadistates/jun2419a/ -out jun2419a_kadi -run_start_time 2019:169:00:00:00.000

./sandbox_starcheck -dir kadistates/mar2519a/ -out mar2519a_kadi -run_start_time '2019:079:22:00:00.000'
./sandbox_starcheck -dir kadistates/mar1119b/ -out mar1119b_kadi -run_start_time '2019:065:18:00:00.000'

And rerun of the weeks and their plots comparing to the pickle temperatures look the same

[jeanconn]

@taldcroft Is this officially good to go now?

[taldcroft]

Excellent!!