Hi, thanks for implementing the dynamic tariff option, really enjoying it!
I got one question regarding “aggressiveness” of the system.
Last week Sunday with plenty solar production, the system wanted to charge from grid.
This week, where solar production is rather low, the system does not plan to charge from grid, neither delay discharge afterwards.
I think it would make sense (17% efficiency factor), battery is at 0% soc, and weather forecast doesn’t look good.
Lowest prices start in 1 hour and right now the plan looks like this:
EDIT: The forecast changed, and the system started charging from the grid, but only to 41%. I would have expected it to charge to 100%, because at the moment a full battery is empty at 2-4am (heatpump). A little more „aggressiveness“ would have been nice.
@merlinste. Welcome to the OpenEMS Community, thanks for getting in touch and for the generally positive feedback.With your name I was able to identify your FEMS-System.
The downside of the “Artificial Intelligence” system we use is, that it is a bit hard to explain why it is behaving the way it does. I have following explanation for this case:
Availability of Day-Ahead-Prices from Tibber is always only from 1 pm, so before 1 pm we can only create a Schedule (“Fahrplan”) till midnight.
Due to the way the algorithm works, it will always target an empty battery by the end of the Schedule
→ before 1 pm it would never schedule to charge for your next-day-morning consumption
Additionally yesterday was a very sunny day (34,4 kWh) vs. only 14,1 kWh today. The prediction algortihm tries to adapt, but most likely it was “expecting” more production.
The development focus till now was a “Winter-Mode” that charges during cheap wind energy hours in the night. We are currently planning evaluating some features that would help in this situation (“Spring-Mode”), e.g. it would have been more efficient to charge the (DC-side) PV production to battery instead of supplying the consumption here.
Thanks for your fast reply (on a Sunday!!).
Yes, and I want to emphasize how happy I am with the Fenecon system. I love it, and I would not say this about many technical objects
My purpose was to give some feedback and room for improvement, as I generally think that this type of intelligence is the future, and I am generally impressed of how good the prediction algorithm works.
Your explanation about the time point helps a lot in understanding the logic.
One idea: I use greenshare.energy (I think they use other data sources themselves) for an estimation of % green energy in the grid over the next 10 days. Another piece of software that works surprisingly well, and the correlation between % green energy and prices is crazy. Helps me a lot in planning further ahead when to expect cheap energy. Maybe data like this could be used to create a „bigger“ roadmap some time in the future.
I did not know about https://greenshare.energy/, but their approach sounds a lot like “Grünstromindex” by our friends from STROMDAO (founding members of OpenEMS Association and member of the board since the beginning):
@zoernert: Let’s discuss if and how we could use your API to improve the price trend beyond day-ahead.
