I realized that as a human I act pretty dofferent than ESS in a special
Situation.
In very cheap phases - e.G. when price per kwh is in the 10% bottom of the usual prices for the last few weeks - I use to fully load the batteries from grid as even when I don‘t know the price after the day ahead there is a high possibility of being pretty cheap anyway.
So my personal Humanplrediction if buying electricity is worth it does not end within the following day.
When prices are vers good I assume that potenitally there will be not a much lower price spot in 48 hours and fill in more than needed for
The upcoming 24 hours.
Is there a possibility to tell DESS to always charge batteries 100% if price is unter a specific point or even „lowest 10% of last week“.
This would be a cool function!
Charging the batteries at particularly favorable times would be one thing. Another thing is to only use this energy when it is more expensive than at the time of purchase. Above all, so expensive that it also covers the conversion losses and battery costs. Only blind charging would result in the battery being completely full and the DESS using the energy until the end of the next price interval. In recent weeks I have often used the cheapest Sundays to charge. Then I set the Min-SOC to 75-80%. Over the course of the week I then release SoC bit by bit to cover the most expensive phases. This has proven to be a good strategy for me.
The element of only usung it when Price + Losses is worth it worked good as far as I see.
The pont is that pretty recurring you have a price point very low here (e.g.21 - 23 cent). But this is usually 2-3 days a week.
The other days prices are in between 28 - 39 cent.
So it absolutely makes sense charging not only to e.g. 60% in a price low but instead to 100% and use the 60% when worth it but have stored low price 40% for the day after tomorrow.
But that is exactly what the DESS would not do in its current configuration. The DESS only calculates the time for which it has prices. For me as a Tibber customer in Germany, that means that on Sunday at 2 p.m. the system takes into account the prices and consumption until midnight on Monday evening. If I calculate with expensive prices beyond that, or rather much more expensive prices, I have to do that manually. I go via the Mind Soc and leave the rest of the decision to the DESS. This has helped me a lot over the past few weeks during the high price periods.
1 Like
I don’t question the fact that the DESS is working really well at the moment. After many ups and downs, I am happy, satisfied… yes, enthusiastic. But your suggestion to buy electricity at particularly good times doesn’t get us anywhere. The system would then use the electricity immediately without taking the purchase flow into account. I once had similar ideas.
To raise the minimum SOC was a good idea as buffer. Today I would have ben runned out of energy 2 hours early if I wouldnt have set 20% buffer with this „trick“.
Now I‘m curious how DESS will act when at 8PM i will at 5% SOC tell him to hold back 25% SOC monimum.
I know with active DESS it does not start immediate charging like in normal ESS mode.
But I‘m curious if it calculates the charge for tonight including the buffer or there will be some error and he ignores the minimum SOC above actual in his forevast calculation.
If it respects the buffer the minimum SOC is a very good tool to fulfill my wish to charge a little more than forecast assumes.
Tomorrog e.g. we will bake cookies and need 1.5kwh more than the DESS can estimate.
Great if I just step up minimum SOC to 15% (which is around 1.5kwh here) this evening.
I can tell you pretty much exactly what happens. If the set min Soc is higher than the actual Soc, DESS will try to compensate for this as quickly as possible. Either immediately or at the latest after the next full hour. Feel free to observe for yourself
1 Like
Yes it immediately charged as Sarowe said.
So i dont see a possibility to tell the DESS with 5-10% SOC charging the tomorrow schedule witz 25% minimum SOC at the moment withoud activating immediate charge which might make nonsense depending on the proce point when you want
To set the percentage up.
sorry, I shouldn’t be right. I’ve tried it exactly like that several times. But you were very sure. Nevertheless, for me it’s still an option to pre-charge some energy on the cheapest Sundays for the most expensive times of the week. It’s just a bit of speculation
From a perspective of risk calculation we need a stochastic model for the day after tomorrow as a step in Victrons DESS.
Last 3 days showed perfect what I mean.
Sunday we have had a cheap price theough the whole day (~21 cent) close to the bottom line prices have ever been in the fall (~19,9 Cent absolute lowest point).
The first jours of monday the price raised up to a normal level between 29 (fall normal low) and 38 (fall normal high).
You already saw that lowest price point at the evening of monday is far above lowest price pijt at the morning.
What the system wanted to do: not knowing the prices for tuesday it figured out just using the 21 Cent bought elecreicity straight from the beginning of monday starting at 3 AM at around 24 Cent so that battery would run out of power close to 11:55 PM
What I did: I thought about the average within the week prices bewenn 29 and 38 Cent and decided not to use the battery unter 30 Cent.
Because my stochastic model tells me spending battery SOC at 25 Cent will probably not be the smartest thing as I assume tueaday lrices again to be in between 29 and 38 Cent.
So I used Grid at 25 - 31 Cent and stored the rest for tiesday.
Which was totally correct.
Today (Tuesday) I see that tomorrow (Wednsdsy) end of day price at 11:55 PM ends lower then usual at 26 Cent so I pretty can spent everything out until that as then a charge price + losses under my average is possible.
@victronEnergy maybe you can implement a stochastic model which switches to a „only price peak discharge“ if assumes a longer period of high prices.
This could be raw mathematical (average prices / average highs / lows per week day) or as well feeded for example with weather data as peices depent a lot on wind forecast and temperature (heating consunption of a country).
I know that Andrwas Schmitz (AI Professor, YT Akkudoktor) already has some model like this for his private free to use DESS Software.
Maybe it might be an idea partnering with him if it would be fine implementing it into you code.