Auto-dampening with lots of cloudy days

[TCWORLD]

I've been using the auto-dampening now for a while, nice feature. In my case there are lots of trees around that cause all sort of shading issues in the winter, so the dampening should hopefully help with predictions for that.

After some playing around to get the right data in - e.g. reimann sum of solar power rather than using the energy tallies (0.1kWh resolution) from the inverter, it was getting quite good at determining the dampening. However there have been a lot of cloudy days recently, and it seems like that has a big impact on the long term picture its building.

Looking at the debug logs, it seems to be a combination of a limited number of days used for calculating factors, combined with long spells of cloudy weather. This is the actual vs predicted for the last month:

So for example it had worked out some reasonable factors , but then there was a solid fortnight of bad weather and it seemed to then lose its dampening - all the factors reverted to 1 as the old data dropped off the end. Notice how by mid feb the predictions jumped massively as the factors vanished.

There were then some clear days around the 16-20 Feb that allowed it to calculate the factors again. The accuracy of the predictions became pretty high again. Same thing happened for most of the rest of february (gray and dull), so at the start of march all the dampening factors reset again.

There have since been a couple of clear days, and its resumed good factors. This is a close up of the last 9 days - notice how in the middle it loses the dampening until the couple of nice clear days (where prediction was way off due to lack of dampening) before regaining:

We're now at a point where it has a reasonable map again for this time of year (later in the year that will improve significantly as the sun rises higher and higher about the tree line):

---

So anyway, on to the questions.

1. I need to check the docs, but I assume it is possible to increase the length of days that it searches back for factors? If so that would help in the winter, but I wonder through spring/autumn, the factors are going to be changing more rapidly, so extending its limits might not be desireable? not sure what affect that has.
2. I'm wondering if there ought to be a way to have it hold on to the "last known" dampening factors when there is a long period without any matching intervals? For example if it finds no matching periods at all, but it previously had some, keep the old dampening.
3. Is there/should there be a way to requesting that once new dampening factors are calculated after a period where none were known, that it goes back in time and re-applies them to a limited number of previous days? That would improve the energy dashboard view of actual vs forecast by removing the big peaks where dampening was lost. Just an aesthetic thing, but would be nice.

[gcoan]

Have a look at the advanced options Tom https://github.com/BJReplay/ha-solcast-solar/blob/main/ADVOPTIONS.md in particular Key: automated_dampening_model_days and automated_dampening_preserve_unmatched_factors

Your question 3 I think won't be done. There's no desire to rewrite history. At the end of the day its only a forecast, but if you make these config changes it'd be less required anyway

[TCWORLD]

Thanks for the speedy reply. Looks like they should do exactly what I need. I agree with those the third question is probably a redundant function.
Seems as ever I should have read the docs more closely 😂.

I presume the dampening factors are stored in one of the various JSON files? If so I might look to see if I can just manually correct the couple of days that the factors were lost for.

[BJReplay]

If so I might look to see if I can just manually correct the couple of days that the factors were lost for.

Make a backup, first.

A typo can ruin your day!

@gcoan is right, no interest in rewriting history if you retrospectively change your mind / read the documentation and change the rules.

[TCWORLD]

Sorted. Relatively simple to tweak the solcast-actuals-dampened.json file to correct the couple of recent days that were affected worst.