Tariff Comparison - Compute the *previous* day?

[WhatIsThisAllAboutSilly]

I cannot see how to make predbat do this, but is there any way to make the comparison function compute the previous day at midnight, rather than predicting?

Its predictions have NEVER been correct, and thus gives the wrong impression on which tariff is actually cheaper. I understand some people use it to automatically switch tariffs but I'm working on the basis of which tariff would have been cheaper over a longer period, as its likely I'll want to be on Cozy six months of the year and Agile the other six months.

I'd like to actually see what the cost would have been with real data, and thus after 7 or 14 days I make the decision to switch.

Thus can predbat be switched to compute the comparison for previous data rather than predictions (Or indeed, do both!)

[gcoan]

You say the predictions are never correct.

The compare function runs the plan evaluation for the specified tariffs based on the predicted load and solar generation so if these are not right for you then that needs to be resolved as otherwise predbat isn't giving you a good plan.

As for your query, that's not a predbat feature that's possible at the moment

[WhatIsThisAllAboutSilly]

The load is an estimate, not actual, and thus depending on what happens that day it can be a good £1 or more out compared to actual.

Examples... one day we go out for dinner then there is ~1-2kWh of load between 5-7 that is missing as we've not used the cooker, lights not on, TV not on.
If I work from home one day there is another 250W for eight hours that may not be there the next.
We might put the dehumidifier on to dry clothes one day, so another few kW that may or may not be there.

So the accuracy of the comparison tab is always going to be wrong unless you do exactly the same thing, at the same time, every day, which is obviously nonsense, otherwise why get predbat to update regularly throughout the day if it knows at midnight exactly whats going to happen.

I can't even back compute it because the logic on if and when the battery charged and how much etc is all down to the actual load, actual SoC, etc.

Without knowing the architecture in detail, but instead of passing the compare function the predicted PV, Load, you just pass the actual PV and loads for the last 24 hours?

[gcoan]

It's an interesting idea doing the comparison retrospectively and I get your point, but I'm not sure doing a comparison on historic PV and load would actually produce a much more meaningful comparison.

The reason is that the predbat plan and in particular predbat mode and battery charging/holding/discharging is based upon predicted PV generation and house load. Predbat has to predict those based upon what it expects your PV and house load to be and make battery decisions based on the tariff costs.

If your PV or house load are much lower or higher than predicted, whilst predbat can and does revise the plan every 10 minutes, it can't go back and charge the battery overnight based on hindsight.

So your charging is (generally) locked in based upon the predictions for the next day and having a comparison done on actual PV and load won't change that, and in fact would produce a less realistic plan. Your best way to get an accurate comparison is to tune your PV and house load predictions; load ML does a better job than the 'classic' days_previous load engine, but of course it still can't deal with random variations on solar and your home activity. That's just the way it is.

[WhatIsThisAllAboutSilly]

Agreed that predbat has to plan based on what it thinks will happen, no way round that it is what it is.

But getting an accurate picture of the cheapest tariff using actual data, would be very useful