The Powerwall has two modes - Time based control and self powered. 

Self powered mode will only charge from the sun - in summer that may be the right choice.

Time based control is supposed to be intelligent - it modifies how much it charges the battery overnight depending on what it's learnt about your consumption and I'm pretty sure on the expected solar production. Time-based control definitely works on our system - it usually does some charging overnight but quite often nowhere near to 100%.

Here's the last few days charging on our system

You can see the overnight charge varying with the changing weather we've had over the last few days. 

Hope that helps

A bit more...

Last summer we ran in self powered mode - once the heat pump isn't heating the battery is plenty big enough to keep us going from day to day with 'moderate' sunshine.

This year thinking of trying an experiment to see what time based control actually does in summer - how little charging will it do? 

Problem with Self powered mode is that if charge our EV overnight it flattens the battery!

Solar PV is feast or famine - in summer there's so much sun your battery is going to be full by midday anyway even if it starts from the 20% reserve. Make sure you are getting some 'smart export guarantee' tariff for your export.

Cheers 

Thank you both, it's been really helpful.  I'm going to have to keep a much closer eye on it for the short term just to see where we are at.  I wasn't aware that the power wall would learn and adapt.

the "time based" learning and adapting is really basic. Thats the one thing I've been really disappointed about, given all the effort Tesla put into AI for their cars, they've done very little with the PW software. the HW, I have no complaints at all.

It mostly tracks how much you used today, and charges by that much overnight. no awareness of high PV generation perhaps being the only reasons usage to went up (maybe you dumped it into car or DHW), ad my biggest complaint, no awareness of prediction of tomorrows PV.  you also can't guarantee that it will "always charge in full" in winter which is usually what you want. it usually will , but not always, and always is what you need. the case mentioned here is the classic - a couple of sunny spring days in a row and it'll stop overnight charging. then you get a miserable cold day and there's nothing in your powerwall.

after getting fed up with this some time back,  I force an "always charge in full" during winter by running an automated script (in home assistant) that sets "backup reserve" to 100% at the time offpeak starts, and set the same "backup reserve" back to 0% when peak starts. That has the effect of forcing a full charge during the offpeak period. been running this for more than a year. the hard part is getting API comms with Telsa as they really don't like talking to 3rd party software. But some far cleverer folks than me have finally figured it out.

shoulder seasons are much harder , I'd like to have something based on "tomorrows PV prediction" to decide how much to charge by overnight, but haven't been able to code such a thing yet.

I then change to self-powered in the summer.

@iancalderbank Time-based improved in a software update last year - the one where you could enter the unit costs for the different time periods.

Before that definitely had the issue with it not fully charging in winter - really irritating - but hasn't happened at all (that I noticed) over this last winter.

Agree that shoulder seasons are tricky - but for me Time-based control has been working ok. I'm going to see how it gets on as we move into summer. Our daily use in winter is pretty high because of our heat pump. 

Good to know the HA integration makes a 'manual' charge regime possible by adjusting the backup reserve. I was thinking of using the same integration to change to time based control when I'm charging the ev (in self powered mode) to prevent it flattening the battery but if TBC behaves sensibly in summer maybe won't be needed. We will see.

Tesla could help by sharing exactly how time-based works - but of course they don't.

@steevjo yes fair point there was the update where they added costs. but I've been running the HA integration to set reserve since before that update so didn't change anything in my setup.

I think with a heat pump in winter its simple = always fill battery. But for shoulder seasons, I'd really like someone clever to come up with a setup to estimate tomorrows heat need and PV generation (Based on the weather forecast, so not impossible) and turn that into an estimated "how much to fill the battery overnight" number.

I've still been running the "always fill" setup these last few weeks and on sunny days that is definitely over-filling, even with the heat pump taking load.

I just wish Tesla would add some extra features to TimeBased for this kind of thing. my suspicion is that as their home market, the US, doesn't do grid charging, they prioritise those functions less.

@iancalderbank It's an interesting problem and far from straightforward. The risk reward balance is massively skewed towards not undercharging as that means unnecessary grid import at 42p kWh for me. I will attempt some back of the envelope calculations for what efficient shoulder charging might save but suspect it's not going to be very much.

The solar prediction side of things is a solved problem. The HA Forecast Solar integration has a sensor available from midnight for forecast estimated energy today. The graph below compares the forecast to our actual solar production over the last week.

You can see it does a good job 6 days out of the 7.

I will a post a link to the money side of things calculations if I get anywhere. Will also attempt to monitor time-based control - have to be able to do better or what's the point!

@steevjo yes indeed. much better to overcharge rather than under. pay 45p/unit if you undercharge vs throwaway some free if you overcharge. I hadn't used that new solar integration , thanks for the pointer to that. I was using something really old self-coded with data from the Solcast API which I gave up on a while back.

I've been monitoring how Time-based Control performs for us over the last 9 days.

Results in this image.

All the details here (bit long for forum post): https://ourhomeelectric.co.uk/is-powerwall-time-based-control-any-good/

The Powerwall never undercharged. On four of the nine days it did overcharge but the total overcharge cost was only £1.08. There might be 100 ‘shoulder’ season days a year so multiply that cost up and it’s only around £12 a year – hardly worth considering. One way of looking at is as an insurance premium against the much more costly undercharging.

The result includes the effects of our behaviour – on sunny days we use the dishwasher, washing machine, heat pump hot water heating during the day rather than at night, surplus goes into the EV if we can. This means our sunny day consumption tends to be high even if the heat pump is off – if we didn’t do this then it’s more than possible that the overcharge number could be (much?) higher.

Going to monitor it again in May and June and see how things change. Would be nice to run TBC year round as it saves having to implement (and maintain) HA workarounds to avoid flattening battery in summer when charging EV overnight. 

Just a quick update - carried on monitoring the powerwall in tbc mode

Last night it did not charge at all - first time I've seen that behaviour.

Sunny day is forecast and heat pump not likely to be needed for heating so probably a good decision.

22 days monitoring so far - never undercharged, overcharged (with hindsight) on 11 days at a total overcharge cost of £2.72