Does Dampening help with low forecasts when thick cloud?

[Javert1969]

From the documentation, it seems like the automatic dampening function is designed to cope with shade and it takes the figures from sunny days?
After running Solcast for a few months, my pattern is that when it’s a sunny clear day, I almost always generate more KWh than predicted by Solcast (often even above the estimate90), but on cloudy days, I seem to nearly always generate less than the estimate - usually a lot less. Often on days with thick cloud I only get just around the estimate10 figure or a bit higher.
My impression from the docs is that auto dampening won’t help with that as it will discard the cloudy days slots?

[gcoan]

First thing to say is that auto dampening will only dampen downwards, doesn't increase forecasts upwards.

If you are consistently exceeding your forecast 90 then it suggests that your Solcast configuration is wrong, check the Azimuth, panel sizes and the panel efficiency. If you get these right then you should be consistently getting aroundabout the PV50 value.
If your settings are correct, try increasing your panel size until you get more accurate forecasts.

Auto dampening uses a comparison of estimated actuals (from satellite observations) to solar forecast to calculate dampening factors based on shading or similar. So yes, it doesn't use cloudy day generation figures.

How many panels do you have on each string on your inverter? It may be that on cloudy days you're not getting enough power generated to start your inverter up. Sometimes the inverter has a startup voltage you can change

[autoSteve]

It sounds most abnormal, yet normal-ish.

I find that with thick cloud the forecast accuracy is marginal. I usually get closer to p10 actual generation.

In a clear-sky forecast the p10/p50/p90 will be close together, and the actual sits pretty much bang on the middle of that convergence, so actual should be close to all of p10, p50 and p90. I call this circumstance a 'confident forecast'.

A visualisation would be handy to describe things. Visualisations can also pretty quickly rule in or out azimuth setting issues.

On azimuth 0=N, +90=W, -90=E, +/-180=S. If you ask ChatGPT it will assume 180=N, which is not the Solcast convention.

There is an example Apex chart in the doco to visualise the forecast today as well as actual generation. It's cool. @gcoan shared it with the community, and I use a variation of that yaml.

[BJReplay]

My Solcast Azimuth and tilt are set as per my installer documentation so if it’s wrong then my installer is also wrong

This is almost certainly the underlying problem.

I don’t see a setting for panel size - the only settings I have available are capacity AC, capacity DC

DC capacity is panel size X number of panels.

At a guess, if your azimuth is correct, your tilt isn't.

But, if you are consistently exceeding 90%, then you have configured Solcast to underestimate your production. On a sunny day, your system's actual production proves that your Solcast configuration is incorrect.

If your tilt is wrong, you would expect that. Especially if your actual tilt matches your ideal tilt, but you have entered a tilt that is quite different. Solcast will underestimate for sunny days, but expect more diffuse isolation on cloudy days.

[Javert1969]

Hmm - ok actually you may be right. My installer original quotations said the tilt angle was 40 degrees, but now I am looking at the post install documentation they have changed it to 35 - so it's not way out but 5 degrees out. I will change my solcast settings to 35 and see what happens, but won't this also have the effect of making the forecast even lower when it's very cloudy?

I guess ideally I need to actually measure the tilt of my roof myself instead of relying on the installation docs.

[BJReplay]

Just set it up accurately!

[gcoan]

I guess ideally I need to actually measure the tilt of my roof myself instead of relying on the installation docs.

plenty of videos on youtube or guides on the internet of how to do this

typically either if you can get in the loft, measure the angle of the roof beams using the spirit level app in your phone

or look at the end gable wall of the house, count the number of bricks across and up in a triangle of the roof slope, then count the same out on bricks you can easily access and measure the angle of the replicated slope

[autoSteve]

My impression from the docs is that auto dampening won’t help with that

Good impression. And if you suspect forecasting issues then turn it off. (Also said in the docs.)

Does Dampening help with low forecasts when thick cloud?

No.