Super weird. Guess I have to fetch my own HP2019 to try and recreate what happens there
Looks great! Is the plan to get flowcards/logic cards with the forecast for tomorrow in kWh? I primarily need tomorrowâs forecast in my flows.
Definitely! Focus mostly has been first to get usable forecasts. Next up is using forecasts in flows.
My end-goal is to make PbtH more a HEMS. Day ahead pricing is already in. Batteries are already in, PV is now in. Next up is EV and home usage prediction (the most difficult one I think).
But this is really still far on the horizon. PV should run stable first.
6 Likes
I like your ambitions â a donation has been made now.
Super! Many thanks for the donation 
The graphs look better today. A reasonable forecast already on HPâ23. On HPâ19, a small bump between 17:00 and 19:00. The output power is being plotted correctly today. So Iâll just put it down to a hiccup for now. It probably has to do with the devices being switched on at the end of the day. In any case, it looks very promising. I am going to remove the summary for the PV. Having two things tracking exactly the same thing is, of course, a bit nonsensical.
@Johan_Bruyn Many Thx for your donation !
I will make a new Virtual Device âSMA Datamanager Mâ. Now not only with power (W) but also energy (kWh). All this for a more correct solar-prediction.
Predicted Learning-problem: Iâm dimming with netto negative prices (0,018-0,018).
Inverterâs output seems to be capped to 600W max?
You could compare that value in the insights graph of the inverter itself, to see if itâs the device or the PbtH graph.
Here theyâre pretty similar:
Question, when will the solar forecast be updated? Midnight? And idea, prevent the value from being lower than 0. With an external kWh meter you also measure the âstand-byâ consumption of the inverter. You will then see the graph dive below 0. Perhaps has a negative impact on the prediction. Another thing, today the value of the PV stuck at 50W while there was no more production.
PbtH solar takes the realtime power data from the source device, just averaged a bit to smoothe out transient spikes.
The power forecast you see is the product of radiation forecast * PV distribution model
The radiation forecast is updated every 15 minutes 24/7, based on a number of weather models. The PV distribution model is trained continuously/real time, plus every night around 1am it will fetch 14 days of Homey insights to calibrate the model.
PbtH takes whatever your source device reports.
1 Like
So the prediction should improve around 1:00? Because right now the graph only shows a peak of 9 Watts between 17-19h on the HPâ19. For tomorrowâs forecast.
Then I think we have a small bug after all. If I overlay the kWh meter and the Solar Forecaster, I get this image.
Reboot both systems?
*edit
Hmm, I found something. On the HPâ19, the power is registered twice: positive and negative. Possibly a result of the beta test. Iâm going to ask Jeroen about it.
Nice! Love the this possibility and tried it. Would it also be possible to use a variable for the time after âandâ? Now it is only possible for the and .
After looking into what might be going wrong with my kWh meter with @Jeroen_Tebbens for a while, and finally removing and reinstalling the kWh meter and the Solar Forecaster, it seems that it is now working as desired. We will just wait and see what happens next.
Quick request. Can you add the âremaining todayâ status indicator?
1 Like
I identified this as an artifact from left-over deprecated power capabilities from older Homey firmware versions. In the next release I ignore those capabilities.
1 Like
Found the issue. Is fixed in next release.
new version 8.7.2 ready for testing: https://homey.app/a/com.gruijter.powerhour/test/
- Improve SolarForecaster source data fetching.
- Fix several flow cards.
- Homey-api 3.17.4
5 Likes
