what could affect the prediction being wrong?
asked
solar forecast is wrong
Various reasons. You need to see if the irradiance forecast vs actual is correct or at least close.
I have found the forecast to be hopeless after sudden transitions from good to bad weather and back again.
This is because the algo leans too heavily on what happened the previous day.
I have raised this before as well. In my region at least, it is interesting but mostly unhelpful.
Forecast.solar has yielded better results for me.
Ultimately it is a forecast, so results may vary.
The source data of the forecast is the number of solar panels, the direction and the angle of inclination, how does the forecast get these data?
how can i fix the source data?
You’re confusing vrm and forecast.solar - 2 different systems.
There is nothing you set apart from location on vrm. Nothing you can change either.
Vrm knows what your production is, when the mppts turn on/off and uses a complex algo to use the base irradiance data (chartable via a widget) to forecast production kwh’s.
You can pull this base data via api and model it yourself, if you are that way inclined.
Calculating weather forecasts is an extremely challenging task, there are not so many compute centers doing that on a global scale, as it needs massive resources to do it properly. And even on global scale, predictions can fail because of the chaotic behaviour of weather.
On top, global weather is not the same as local weather. This is even more problematic. Especially some local fogs or whatever effects are not covered by such forecasts, which is especially the problem when weather changes from one stable "good" situation to a "bad" situation. There you will have a lot of local effects not properly covered by any forecast. It might be even a super sunny day forecasted and there is just one little cloud traveling through, which is then just noticed by people living on this "line of travel" of this single cloud. But the local weather forecast is still a super sunny day with <1% clouds or so...
Especially in my region (in the central alps of Europe) the local weather rarely matches even the forecast made for our country. A 50% correctness in your example I would consider as not too bad in my situation. But that's why I would never pay for such a weather service - it is simply not worth it, especially as private person. I think it is just a clever business model to earn money without giving any guarantees for the data they sell or explaining in detail how and from where they generate this forecasts. At least I have not heard that people get their money back, if the forecast is significantly (and proven) off all the time... ;-)
It might be better, if one lives in the more flat areas, though. But even then, as it is well known from the last few years, that massive rainfalls in quite small spots with devistating outcome threatened several quite small regions, where no one could really tell where and when it will happen, until it was too late.
And here we talk about just a bit more or less sun on some PV panels on an incredible small area in relation to the area resolution of such weather models of the size of cities and over about just 1/3 the time period of a whole day especially in winter season...
The forecast is wrong & I think that the reason is that the skyfilter factor is not correct implemented in the algoritme. I noticed recently cloudy rather correct ; rain overforecast ; sun underforecast. I made a account on solcast just to compair. These data where rather correct.
