-
|
The advanced dampening adaptation feature looks super-cool and I've been wanting to try it out, but can't seem to get it working. I figure it's most likely I just configured something wrong, but I can't for the life of me figure out what. First, here's my I've had the dampening adaptation option turned on for well over a week at this point, and it's been crazy sunny where I am, so I figure that should be enough to pass the default minimum of 3 days of data, right? But this is what I get in my logs shortly after midnight: Here's my On v4.5.1 Any ideas on what I can fix? Thanks for any and all help! P.S., is there any way to trigger the dampening adaptation code to run other than after midnight when estimated actuals are retrieved? An integration reload doesn't seem to do so. |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 19 replies
-
|
Fiddle with this in
|
Beta Was this translation helpful? Give feedback.
-
|
Note that autoSteve is in and I will soon be in the land of the rising sun, so service may be delayed. We're in a similar time zone, but far more distracted than usual. |
Beta Was this translation helpful? Give feedback.
-
|
Looks like you're missing data for 11th March in the json which shouldn't be an issue but.... I'll investigate that later. So you could try editing the json to start at 12th. |
Beta Was this translation helpful? Give feedback.
-
|
Yep for some reason I thought that fix had snuck in to 4.5.0 |
Beta Was this translation helpful? Give feedback.
-
|
@Nilogax @autoSteve I'm already on 4.5.1, but since the default for minimum days should be 3, shoudln't data from that far back not matter? I'll backup and try editing the json when the 429 storm is over. |
Beta Was this translation helpful? Give feedback.
-
|
You're right, it shouldn't be a problem but looks like you've found a bug. Somehow your json has only one dampening history entry for 11/3 - the first block in the file for Model 0, Delta -1. There should be 12, going through Models 0 to 3 and Deltas -1 to 1. I would love to know how the heck that happened, did you have an impeccably timed post midnight outage in the small hours of 12th March? If I replicate that data pattern in my file I get the same error and adaptive dampening fails. If I delete the one and only entry for 11/3 then it all works fine and I get the expected logging below (plus full adapative dampening logging after). So the short term fix is to just crop that one entry for 11/3 from your json (take a backup first), the proper fix will no doubt be in 4.5.2. |
Beta Was this translation helpful? Give feedback.
-
|
Or if you fancy beta testing a potential fix (pending @autoSteve's review) - replace the method _find_earliest_common_history in dampen_adapt.py (around line 636) with this. You'll need to restart HA afterwards. And take a backup of your old file just in case. And wait for the 429 storm to finish. And turn around 3 times anti clockwise while waving some magic seaweed.... def _find_earliest_common_history(self, min_days: int) -> dt | None:
"""Find earliest date where continuous dampening history is available for all models and deltas.
When all model/delta combinations share the same history dates (i.e., no new strategy was
added mid-way through), symmetric gaps caused by missed actuals are tolerated: the date
set is the intersection of all period lists and gaps are skipped rather than treated as
a continuity break (provided enough usable days still remain).
Returns:
Earliest common date with sufficient usable history, or None if insufficient history exists.
"""
period_lists = []
for model in range(
ADVANCED_OPTIONS[ADVANCED_AUTOMATED_DAMPENING_MODEL][MINIMUM],
ADVANCED_OPTIONS[ADVANCED_AUTOMATED_DAMPENING_MODEL][MAXIMUM] + 1,
):
for delta in range(
ADVANCED_OPTIONS[ADVANCED_AUTOMATED_DAMPENING_DELTA_ADJUSTMENT_MODEL][MINIMUM_EXTENDED],
ADVANCED_OPTIONS[ADVANCED_AUTOMATED_DAMPENING_DELTA_ADJUSTMENT_MODEL][MAXIMUM] + 1,
):
if self._should_skip_model_delta(model, delta, min_days)[0]:
continue
period_lists.append(
sorted(entry["period_start"] for entry in self.dampening.auto_factors_history[model][delta])
)
if len(period_lists) == 0:
return None
# Dates present in every model/delta combination.
common_periods = sorted(set.intersection(*(set(pl) for pl in period_lists)))
if not common_periods:
return None
# Determine whether all lists are uniform (same dates, no new strategy was introduced).
all_equal = all(pl == period_lists[0] for pl in period_lists)
if not all_equal:
# Non-uniform: search for the earliest common date from which all lists are strictly continuous.
for candidate in common_periods:
if all(
all(
curr == prev + timedelta(days=1)
for prev, curr in pairwise(sorted(p for p in periods if p >= candidate))
)
for periods in period_lists
):
_LOGGER.debug(
"Earliest common dampening history (non-uniform) is %s",
candidate.strftime(DT_DATE_FORMAT_UTC),
)
return candidate
return None
earliest_common = common_periods[0]
_LOGGER.debug(
"Earliest common dampening history is %s with %d usable days (gaps tolerated)",
earliest_common.strftime(DT_DATE_FORMAT_UTC),
len(common_periods),
)
return earliest_common |
Beta Was this translation helpful? Give feedback.
-
|
weirdly, if I do set The KeyError is actually for the 2026-03-21, which would be yesterday for me. I'll try replacing |
Beta Was this translation helpful? Give feedback.
-
|
Actually, I'm going to leave |
Beta Was this translation helpful? Give feedback.
-
|
That exception is issue #483, @rcode6. Fixed in a PR yet to be merged and released. It's benign unless you force the development flag which will prevent the integration loading. That PR also tolerates gaps in history. My advice is getting ahead of the released code, because land of the rising sun, so go with @Nilogax's suggest. |
Beta Was this translation helpful? Give feedback.
-
|
righto, I'll run the updated |
Beta Was this translation helpful? Give feedback.
-
|
...or A tweak in v4.5.1, designed to tolerate just this kind of thing. |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
@rcode6, you could upgrade now, but be very careful about what will result while you are upgrading, what your expectations will be. The golden rule is to change nothing at solcast.com because a Check list.
Check-list complete. All ticked? Upgrade. Nothing bad will happen unless something material at solcast.com has been changed. We need API. And in a |
Beta Was this translation helpful? Give feedback.
-
|
@autoSteve I neglected to put all the usual info since I didn't submit as a bug report, but this has all been on the latest version, 4.5.1
|
Beta Was this translation helpful? Give feedback.
Or if you fancy beta testing a potential fix (pending @autoSteve's review) - replace the method _find_earliest_common_history in dampen_adapt.py (around line 636) with this. You'll need to restart HA afterwards. And take a backup of your old file just in case. And wait for the 429 storm to finish. And turn around 3 times anti clockwise while waving some magic seaweed....