This is completely irrelevant as working for others does not fix this issue that is obviously there.
What if I would switch off the north directed part of my PV system - do you really think the problem is gone then?
are you reading what I am writing? obviously not.
I asked for someone with insights and doing a root analysis to fix this issue. I do not care about the algo as such (influencing factor of course would be helpful) - just fix it or explain what I should change.
I read what you write. Of course. Otherwise I wouldnât reply. What Iâm wondering is why it works for many people but not for you. So the first thing you look at are the parameters. The only parameters that can be influenced are the location and the operating time of the system. You explicitly stated that both of these are fine. Then the next clue: what criterion distinguishes your system from the majority of other systems. This led me to suspect the northern system. But you seem to have ruled it out and not accepted it. Have I described the status quo even remotely correctly? So why is it that something that often (very often?) works well is so wrong for you? Can you explain it to me?
Why should I explain something only victron knows?
Why do you think if something is working for many the system cannot cause issues for others?
If I have a corner case fine - lets sort it out and improve the system.
I would kindly ask you to stop posting within this thread until you can provide facts instead of believing in something - thank you.
Iâm not 100% sure and can only guess here that either Solcast or Victron themselves are estimating parameters like Peak power and panel orientation based on your live data.
But I agree: you need support from a Victron engineer like @dfaber or @dognose to uncover the root cause.
And surely itâs not your task as a end user to debug this. But still I find it important to acknowledge that VRM+DESS is for free and Victron is driven by community and open source to a great extend so itâs natural that they donât have enough resources to debug a very rare issue like you have.
You just have to hope for mercy here - and add your VRM account number (the number in address bar marked with # here: https://vrm.victronenergy.com/installation/######/dashboard) to your profile description or mention it in this thread here so that support techs can inspect your system more closely.
First of all, I go to church to believe, not to a forum for PV technology. Second, Iâve been here for a while and wanted to help. Maybe even help people understand. However, I have absolutely no problem opting out, which Iâll do with that. Good luck.
I absolutely agree- overall it is a great system - absolutely happy with it.
Still even if happy I am hunting for further improvement 
Peak power could be done as an average and is of course itself also depending on angles and even on dirt and shadows influencing. But lets stop guessing I am willing to provide any data needed.
Iâm not aware of the exact algorithm, but angle or Peak Power of the panels is not needed for Forecasting. It actually works quite easy, by matching your historical production data with the solar iridiance forecasts of any day in the past.
It kinda learns: âWhen the iridiance forecast said 9.0, your system - whatever layout that is - produced 3400 W, so today [from 9 to 9:15], we gonna see 3400W as well, as 9.0 are forecasted for that intervalâ
Thatâs why it needs about 4 weeks of historical data to work accurate.
Thanks @dognose for looking into it.
OK more or less taking the average (rolling window of 4 weeks) of matching (irradiance of the hour as criteria) days of the past. Should work ok for any system no matter what angles or kwp it has even with multiple connected inverters this should work.
Comming back to the issue I see:
The forcast jumps from extremly unrealistic values to somehow matching ones within hours (without a changing irradiance forecast).
As this was not getting better even after several month, I resetted my historical data ~one month ago. But still we have this jumps.
E.g. just now (at the evening) it shows for next day
So the question is where is this super unrealistic (>130kwh) forecast comming from? THis is initially causing wrong battery charging plan - luckily in most cases corrected in the morning of the next day.
Update on next morning:
Maybe also worth to look into is the mid term forecast which shows the issue also very well:
update after lunch - today it does not seem to correct itself:
and for tomorrow:
How many inverters and strings do you have @endurance and how is their data getting into VRM?
Just a wild guess, but is it possible that your momentaneous power reading (possibly aggregated from multiple strings / invertes) has tiny spikes that confuse the peak power detection of the forecast?
5 Inverters. 4 directly in Venus OS (SMA IP connected) one (but only 250W) via mqqt-dbus adapter (also all local IP).
At least in all data (logging in 10s intervals) I have, there are no spikes visible.
Spikes and peak power usage could be a reason but if I understood the way it works correctly there is no peak power used but rolling 4 week âaverageâ energy per hour matching irradiance.
What is bit irritating is the switch between pessimistic/realitic and extremly high forecast. If always a 4 week period average is used there should always similar results at least not big changes (unless irradiance changes).
system ID was provided - hopefully someone from dev team is looking into it.
So far it was not that bad, as it corrected itself regularly. Means that battery was still charged if price (typically around lunch) is far lower then the daily peaks (in the morning and evening). But now it seems to stick at the high forecast not planning to charge from grid additionally
I will have to force the system to charge then via external control.
Few hours later even worse upto 400kWh in one day
Maybe one more observation which I just spotted
It is one screenshot from VRM. Schizophrenic system?
Yesterday the forecast looked OK the first time and today for tomorrow makes also somehow sense:
Also mid term
Seems the team found the issue and hopefully fixed it permanently.
or the algorithm learned without the programmers changing anything.
Jetzt kommst Du schon wieder damit.
So ganz urplötzlich lernt das System - was ist denn wahrscheinlicher, dass das System etwas innerhalb von wenigen Tagen nachdem ich Kontakt mit jemanden aus dem Support hatte etwas lernt (was es in einem halben Jahr nicht gelernt hat), oder das es einen Bugfix gab? Bugfix kann auch das beheben von falschen DatenbestÀnden etc. sein - das weiss nur Victron.
Wir können uns drauf einigen, das es hat gelernt einen (aus meiner Sicht) âsimplenâ Forecast ârichtigâ zu berechnen.
Aber nicht den Tag vor dem Abend loben - die Vorhersage ist erst zwei Tage halbwegs korrekt.
It does not really matter what was changed - if the result is improved (I assume also for others with similar issues). cross fingers it keeps running fine.
I think the thread can be now âclosedâ
forecast is still fine
and support confirmed that they fixed something on VRM side (several users were impacted - but not many). Means my issue was not an isolated local config one, also not depending on directions of the PV modules (north facing) etc.
Thatâs probably a bit simplistic. My forecast isnât working as well since the forecast period has been limited.
The issue which I described (for me) where forecast is off by factors - seems to be sorted. That the forecast deviats several percentage I do not mind.
Please open a new topic with the concrete issue and some background facts. If done already you may add the link here so others can find it easily.
long ago and contact exists
(post deleted by author)
