Calculation "Remaining day"

[felixheyne]

Hello!

Solcast allows 10 API calls in the free subscription. These calls are usually recognizable by changes in the forecast values.
So far, so good.
If you now look at the curve for the "remaining day" value, it seems that values ​​are being generated every 5 seconds; a nice "curve" is forming.
Where does this calculation come from and how is it performed? These values ​​can't be coming from the API.
Does anyone have any ideas?

(Thx to google translator...)

[BJReplay]

Good question!

That is a gap in the documentation!

The remaining sensors (including the this hour, next hour, next X hours, remaining today) base the remaining amount from an interpolation curve.

The interpolation is very briefly mentioned in the readme, but only as an aside.

The primary reason is to smooth the curve.

The remaining forecast is smoothed into a curve in solcastapi.py - look for def_spline

[autoSteve]

To correct you slightly, @felixheyne (and it might be a translation thing), the values are interpolated (and updated) every five minutes, and not seconds.

Historically (before BJ and I became custodians of the code) this value was interpolated for many moons. When we took over it was a simple linear interpolation, or followed a straight line between per-half hour intervals. While good, this was visually jarring when historical data was charted using something like an Apex chart.

I implemented a much more sophisticated 'spline' interpolation later, which follows an interpolated curve to more closely match what the sun does.

In fact, two distinct types of splines are built. One set for the "remaining" sensors, and another set for the "momentary" sensors. If 'estimate 10', 'estimate 50' and 'estimate 90' are all enabled then there will be a total of six splines built at each forecast update or on startup.

For the "momentary" sensors you can see this at work by examining and comparing 'Forecast next hour' (enabled by default) with 'Forecast next X hours' (disabled by default). The 'Forecast next hour' is not interpolated. 'Forecast next X hours' is, and when set to an 'X' of 1 will show a different value to the next hour sensor. The reason is that the 'X' hour sensor is from now for the next hour (i.e. from 10:40 to 11:40), while the other sensor is literally next hour on hour boundaries (i.e. 10:00 to 11:00).

[felixheyne]

At first: seconds was wrong by me, minutes is correct.
Thank you very much for the enlightening statements. That helps me a lot.
Good job!