DESS next-day schedule based on simple price prediction

My electricity prices (Tibber) for the next day usually get updated on 13:00 so VRM will not be able to create a schedule for the next day until then.

But sometimes VRM seems to take the price of the last hour (23:00) of the current day as a “prediction” for the prices next day and creates a schedule for the next day also (which gets updated as soon as the real prices are published on 13:00):

e.g today it will be 29,98 Cents/kWh on 23:00 and for tomorrow it assumes 29,98 Cents for the whole day:


Strangely enough, this very primitive price prediction does not happen every day! Sometimes there is just no schedule for the next day until 13:00 when the real prices are published.

And this is puzzling me. Is this “prediction” really supposed to be a prediction or is it just a bug???

And if it’s supposed to be a prediction (which I hope) then I would suggest a even better prediction strategy:

Just assume the price will be the same as today but on a hourly basis. In most cases this will be a good prediction. In a week full of sun prices are low around noon - which is likely to happen on the next day too.

One way to make that a bit smarter would be to perform this lookup-table kind of prediction work on a weekday-schedule, i.e. hourly prices of tomorrow are most likely the same as those on the same weekday last week. This will automatically account for usually lower prices on weekends.

But of course there also will be occasions there the prediction is off. E.g. when the weather has changed a lot or when there are public holidays. It cannot be done perfectly. But if you are afraid of user complaints you can just make these price-prediction strategies an option for advanced users and include a disclaimer…

I’m observing exactly the same thing. I know the Victron team there is working on some things and conducting some tests. I find this development extremely interesting and am watching it with great interest. Let’s give you some time to see what comes out of it.

In my view, the forecast price should be based on past weekday prices and a rough grid of weather data (sun and wind). Another approach might be to access yet another index of stock market data. I know that there are other indices besides the day-ahead price, but I don’t know exactly which of these indexes reflects the price level for the coming week. In my view, the accuracy of the price isn’t that important; what’s much more crucial is that the system recognizes weather and consumption data over a longer period of time, thus identifying PV undercoverage earlier and thus using PV energy more effectively. But these are my own thoughts on the topic.