Now that I have been running my own version of this software concept, I noticed it’s somewhat suboptimal at times.
Say I don’t heat during the n most expensive hours, and one of those n is at 23:00 (which is not uncommon) so I don’t heat by then.
But now what if next day prices are overall higher than today’s … I will start heating at 0:00 because that’s among 24 - n cheapest of the next day but that hour it is more expensive than it was the hour before so reversing those would have been the cheaper choice in total.
Ok we don’t obtain prices more than 1 day in advance, but by 13:00 or so they are there.
I now came to think that instead of doing the calculation on a static 0:00-23:00 basis, we should better be using a sliding window of the next 24 hours (or as few or as many as we have prices available).
If looking to cover a fixed period of 24h, we might need to estimate prices, say it’s 12:00 and we haven’t received next day price info yet, we would need to generate a value to calculate with for the hours 13 (= 00:00 next day) to 24.
wdyt?