Hi @Barbara ,
Did you read my recent contribution VRM PV forecast (improvement appreciated) - Comparison Victron's Solcast vs. native Solcast - #16 ?
My analysis clearly shows that you potentially add and modify the data too much.
The solar forecast from VRM is simply not reliable anymore after modifications.
I deal a lot with weather forecasts both professionally and privately, so I am aware of the complexity. A decisive factor is how detailed and regional the weather data is. What I am always impressed by is how well and quickly DESS reacts to changing forecasts. I also have several weather reports and forecast programs running here in parallel. Some of them are from the paid professional sector. The VRM forecast is actually very, very good.
Hi Thomas, yes I read that, my point being is that there is no ‘Solcast solar forecast’: Solcast forecasts irradiance, and VRM uses that forecast to forecast solar yield.
Solcast has no notion of what solar panels produce with a given irradiance, so just using solar irradiance would make Dynamic ESS useless (we tried this as well in Beta).
So I agree with you that forecasting solar is less accurate than forecasting irradiance, but I am also saying that to close the gap in accuracy completely would mean that we control all factors in the model, and that just is not realistic. We are just continuously looking for ways to decrease the gap.
Hi Barbara,
in case you use Solcast’s “Get Irradiance and Weather Forecast” API, I see that you making the challenge larger than necessary.
Why don’t you use the “Get Rooftop PV Power Forecast” API? In that case you overcome all issues that you struggle with.
Recent days weather and weather forecast was more than stable. Anyhow your 24h rolling PV-forecast remains anything than stable, whereas Solcast’s was!
@weichi is right, just try Solcasts “Rooftop PV Power Forecast”. There you can parameterize direction, horizontal angle and size of the panels.
How would VRM know what each users configuration is?
The purpose is to use metrics VRM has access to, to calculate a yield, by tracking MPPT state, typical generation given known irradiance etc.
Ok, that’s correct, indeed!
