How the formula for the total energy should look like

[michal-bartak]

Hello,
I have no dumping feature enabled

I observed, however, that total daily energy provided by sensors like solcast_pv_forecast_forecast_today, does not match the sum of partial energy calculated from pv_estimate. (Yes, I know the estimate represents power, not energy)

Question 1: Does any deterministic function exist that describes the relationship between partials and total daily PV energy?

Question 2: What can I squeeze from available data (percentiles, spread) to calculate hourly (half-hourly) energy with the best precision possible? Currently, I use 50% percentile. Is there any room for improvement? ChatGPT proposed as follows:

slightly conservative:
E ≈ 0.7 * P50 + 0.3 * P10
slightly optimistic:
E ≈ 0.7 * P50 + 0.3 * P90

Is it all I can achieve?

TBH, I'm not following the idea of providing power instead of energy. It can get different shapes over time. If it's average during the time period, providing energy would have the same meaning. Maybe you know better what the justification and benefits of it.

With regards

[BJReplay]

See the answer in #237 for converting power to energy.

[autoSteve]

Average power. Ask Solcast, not us. It is all we have. Raise a support case on Solcast and ask them why they do this. I know why.

We are not Solcast. We are a couple of guys sparing our spare time to make this HA integration great.

Your question? Your answer. We provide a forecast that is provided by Solcast in Home Assistant, and not a battery charge optimisation solution. You might want to look at PredBat. Ask @gcoan. (Feel free to chime in, mate.) PredBat can get fed by this integration.

[BJReplay]

Using these data in the power domain does not make sense

It clearly does for Solcast, because they have built a successful business forecasting the power output of solar systems.

As @autoSteve has already said, any issue that you have with the units that they have chosen to supply and the reason for their choice is one that you can only take up with them.

As the units are able to be converted, we should not care too much about what units they supply.

I am interested both in energy and power; insufficient power means that I have insufficient excess power to charge my EV - so I am interested in the power curve, as well as the energy implied by the area under the curve.

[autoSteve]

As an aside, I use PV50 data only to determine how much of my "usual" overnight use before the sun comes up can be shed as excess charge to the grid, and this is informed by some automations and by manual, "is the GF going to be here?" settings. I use this to determine how much energy I can "shed" to earn premium feed-in $ at peak time for the evening at hand based on morning forecast.

It works well. I do not pay for electricity.

But it is all roll-my-own automations, based on the data from the Solcast integration. A "poor man's" PredBat, probably.

[gcoan]

Your question? Your answer. We provide a forecast that is provided by Solcast in Home Assistant, and not a battery charge optimisation solution. You might want to look at PredBat. Ask @gcoan. (Feel free to chime in, mate.) PredBat can get fed by this integration.

Thanks for the callout @autoSteve predbat could be your friend if you want a pre-packaged solution to automating your battery charging based on energy rates, predicted solar, predicted load, predicted outside temperature (useful for electricity based heating such as heat pump) https://springfall2008.github.io/batpred/

In terms of 'how to use the forecasts provided', Predbat uses PV50 x 0.85 + PV10 * 0.15 as being a slightly more pessimistic. I personally run with PV 10 * 0.25. In my own experience I find that I usually achieve between the PV10 and PV50. Sometimes I exceed the PV50 but very rarely will I get the PV90.

[michal-bartak]

Thank you @autoSteve and @gcoan those are answers I would expect from the Solcast community.
It seems that the Batpred does not support Wattsonic/Solinteg yet. Regardless of that, I developed a simple HA automation that optimizes ROI by charging/discharging the battery at optimal moments, based on an empirical model. I'm about to publish it on the HA forum. Here is a sneak peak

[gcoan]

Batpred does not support Wattsonic/Solinteg yet

Predbat can support pretty much any inverter, it just needs a custom inverter config developing for it to tell predbat what the sensors are and what controls are to be manipulated for charging, discharging, etc.

All the instructions for doing this are in the predbat documentation, I guess no-one with that inverter has used predbat yet, or if they have, they haven't contributed it back to the community.

Predbat isn't for everyone, it is a quite powerful and this quite complex. Good luck with your automation, I was doing similar before I started on predbat.

Cheers

[BJReplay]

It seems that the Batpred does not support Wattsonic/Solinteg yet.

EVCC appears to support Wattsonic/Solinteg and also integrates with Solcast (and integration with HA) so may be worth investigating for a well supported package.