Withdrawal of tuning


I learned from a third party forum that tuning was being withdrawn. I heard nothing directly from solcast by email for example, although the website confirms that this is so. What do forum members think of the withdrawal of tuning?


Well, there is no such thing as a free lunch.

As far as I understood, the new pvlib algo will create better results without additional tuning. However, I think, the rooftop account will simply not allow for complex modeling of the solar array.

For my part, as I have a split array on my rooftop, I’ll take their offer to create a second account.
Based on that, I will adjust my config and split these into the two arrays…fetching data and re-combine them in my local database.
Let’s see if this will work out…


What’s a “pvlib algo”? Is there documentation? I’ve had no communication from solcast.


I just had my first communication from Solcast in response to a ticket. Apparently existing users retain their tuned models.


I received an Info, when I logged into my rooftop account via the push message/operator chat…pointing to this Article: https://articles.solcast.com.au/en/articles/4945263-pv-tuning-discontinued

There is a link to the new pvlib toolkit the have / will be having in place: https://pvlib-python.readthedocs.io/en/stable/

And a reference to the original invention can be found here, I think: https://forecasting.energy.arizona.edu/media/papers/pvlib_fx_pvsc_44.pdf


…good to know. Mine was last tuned, according to the rooftop account in November last year…still working well.
But I am planning to extend my site with an additional, third array soon.


So it seems that SolCast didn’t reach out to you (for example by email) that they’d withdrawn the service either. Instead you stumbled on the announcement on their website. The replacement Python service seems to be a scientific research tool and thus significantly different in nature to the automatic tuning product for home hobbyists that’s been withdrawn.

Or have I missed the one or two lines of killer code that reproduces or improves upon the previous automatic tuning for a home hobbyist?


As fas as I understand, the new scientific pvlib would allow for a much more complex modeling of the array, hence a tuning feature is no longer necessary.
I look at it that way: pv tuning is compensating for the lack of ability to model the complexity of the array (including converter models and such) layout in the old “system” at solcast.
At the end, this is the only factor, besides solar radiation/energy based on geo-coordinates and weather which is still necessary and which will of course remain required.


The new toolset might allow you to build a better solution, equally you might invest a considerable amount of time and effort but have a worse solution. It’s a bit like rather than supplying a finished product like a car you get a supplied a selection of raw materials and machine tools described as an upgrade - you might indeed build yourself a better car but you might not.


…agreed…that is also possible.
However, we can only speculate on the reasons for the change.


ok., I’m also a little disappointed, and the communication was probably not a master piece … that said:

IMHO, tuning had three purposes:

  1. split-array handling - they have the generous offer to allow you to have two sites, which gives you probably a better result than a tuned single-site array
  2. shading - because of (close by) trees or (far away) mountains. I neither have a clue whether the old tuning algorithm would have handled that, nor how to do that otherwise. It would need an awareness of date (sun rises/sets at different locations throughout the year
  3. general efficiency scaling - ie., panels loosing efficiency over time. That’s probably not a very big concern, but one could manually change peak power.

As far as I can see, rooftop sites only support kW based downloads, so … not much need for pvlib anyway; that’s probably different for other account types. But even with pvlib, a quick browsing didn’t reveal an easy handling of fixed shadings due to trees and mountains.

Bottom line: Not much is lost :slight_smile:

What I miss a bit is the limitation to 20 downloads/day … SolCast updates forecasts every 30min or so. I’d even pay a little (but not $99 as is their cheapest offer!) for that.



My solar array doesn’t suffer particularly from any of these issues, but nevertheless tuning turned these pre-tuned rsults into something almost exactly line-on-line.

While the new toolset provides more flexibility for those with time on their hands, the old service did the job well for me and I regard moving from providing the service to recommending tools for you to make the service yourself as retrograde.

In my own case for example the controller that uses the solcast service only keeps 24 hours forecast of a limited amount of future data and only at half-hourly intervals. To reproduce solcast’s tuning would presumably require a major rewrite to maintain a history of both output and forecasts at higher resolution over a much longer period of time even before actually trying to tune in an automated manner.


Interesting picture - the capping in forecasts seems to suggest a strange configuration


Either you are in a really sunny location, so that panels produce more than their name-plate power, or you have AC < DC, causing the capping?


The capping isn’t real and went away after tuning.

The DC capability is indeed a little more than the AC capability. In the UK you need permission to have an inverter of more than 16 Amps (3.68 kW) but I have 4kWp of panels. The configuration is very standard - 16 panels arranged in a rectangle on a roof orientated only a few degrees from south.