Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

The important missing control for Tesla Powerwalls

This site may earn commission on affiliate links.
I’ve had Powerwalls for four years now. As my thread title suggests, from experience, Powerwalls lack one very important, if not essential control; the ability to fill it to a given capacity. I’m pretty sure I’m not alone in the view given that many owners use the popular cheap but short off-peak tariffs which we have in the U.K.
I‘m fortunate in as much as from late spring to late autumn , I have sufficient solar capacity to keep mine filled up. I know that in the depths of winter I will need to fill up most nights from the grid. In the these two scenarios they perform pretty much on the money. Others probably struggle throughout the year.

However, late autumn and early spring have always been a nightmare. I had hoped that major changes to the Tesla app which included adding tariffs and reducing the operating modes to just self-powered and time based control would resolve the problems. But it didn’t. I’ve just switched to time based control for the winter and I’m stuck in the same annoying rut that happens every year!

Recently we in the U.K. have had the usual dull wet days and bright sunny days. Apart from being unusually warm. Autumn has been what it always is and the daylight hours get shorter.

In Time based control, Powerwalls depend upon their AI to work out how much of their power came from solar, how much they discharged and how much they need from the grid. That works reasonably in stable weather conditions. It fails abysmally when a few sunny days are followed by a dull day or two and vice versa. Furthermore, the AI has no idea that even if there is some sun coming, I want my Powerwalls topped up from the grid because I’ll be charging two EV’s from solar tomorrow.

Here’s the annoying part; on two separate occasions last week there was little solar, the first night PW’s were at about 50% at bedtime and one car was empty. I set my reserve to 47% knowing full well that I would have to wake myself up before 2.30am to check what was happening. I did and sure enough, the car had charged but the Powerwalls had only topped up to 62%. I did what I always do and upped the reserve to 75% as I knew that the following day would roughly produce enough to finish the job, which it did. Two days later, same thing. Set a higher reserve at bedtime, get up in the middle of the night to up the reserve again to ensure that they are full enough for the day.

I won’t set a final reserve at bedtime because then I’ll be buying peak rate power and I shouldn’t need to. No matter how good the AI is, it will never be better than me because it can never have my overview of my needs. The answer is simple; an optional slider control to set a charge level exactly the same as my Model 3. Why on on earth was that ever omitted? Perhaps it’s a USA thing. It’s obvious to me and I would get a good night’s sleep!
Does anyone else want a charge control added?
 
Last edited:
I agree @Dilly and I have exactly the same problem. Sometimes my Powerwall has a moderate rate of charge at the end of the day with a 20% reserve set, but under these circumstances, the Powerwalls will not charge up enough in the cheap 4hr Octopus GO period. I therefore set my alarm clock for 1am and then I set the reserve to 75% to force the Powerwalls to charge up more.

But then, during the day, the Powerwalls would drop their charge to below 75%, so before this happens, I need to set the reserve back to say 20%. Then by the evening, I need to set the alarm clock again.

I have this issue in Autumn and Spring when the AI seems less able to cope.

I think I agree with your solution, what I'd ideally like is when Time Based Control is Set, and Grid Charging is set to YES, an extra control pops up with AUTO or a slider in say 5% steps that sets the Maximum PowerWall % Charge - yes similar to charging my Model 3 too.

What do others think, would you find this useful? I certainly would, I'd getter a better night's kip.
 
I've just kept my PW in Time-based control since we got on the Go tariff at the and of Jan this year. Initially the "AI" was utterly useless, behaving as described above. However, since April it has behaved extremely well - correctly predicting sunny/dull days and charging appropriately. Either Tesla have improved the AI, or it needed a few months worth of data to inform its behaviour.

There have been a few squeaky-bum days where I doubted it would do the right thing, but it hasn't failed yet (since April anyway).
 
  • Informative
Reactions: ACarneiro
Here is an example of where my PWs charged to 87% with a 5% reserve in Time Based Control.

We just got away with staying on battery and avoid peak rate. If I had set the reserve at 20% I’d have been charging the PWs during peak cost time.

I’d like this proposed extra control to charge my PWs to 95%.
3BD02B4C-A2E4-4360-A163-1BBA3F215E8E.png
 
One case in point is last Friday morning where I set the reserve at about 48% at bedtime to avoid peak grid draw. I was up about 1.15am And the PW’s were happily ticking over with no real prospect to charge much beyond 60% needing me to up the reserve to avoid running out on Saturday.
762A22BD-0A25-431D-B463-871023E8D0DF.png
 
Totally agree mine is on tbc but I up my reserve to the max I can at bed time just leaving enough to get me to the cheap rate ( go faster 2.30 / 6.30 ) I get up around 5.30 / 6.00 so first thing check app if not at 100% move slider to 100% then at 6.28 move slider back to 5% .If it is a sunny day just top up hot water / put electric towel rails on / leave either car on eco plus just drawing 1.4 unless there is plenty of sun then zappi will just track it . As the day goes by I just do the reverse turning everything of and keeping a close eye on the powerwall % and not letting it drop below 80% .If every thing is maxed out I have to live with some being exported but cheaper than day rate . I was on go 12.30 / 4.30 until December last year getting up twice per night. It has worked very well only occasionally like today got up 5.45 powerwall on 84% so set slider to 100% back to 5% at 6.15 .But could have a Lie in if what p Dilly say’s happened making the powerwall the no 1 battery by far
 
I'm 100% with Dilly also.

It drives me mad at this time of the year that on Time-Based control the AI in the PW will sense the fact that on one day "we've used a lot of electricity" - namely my Eddi has converted all the solar spare to hot water and think "I must charge myself to 100% in case that happens tomorrow" - and then next day there's next to no spare capacity in the battery, I've paid for all the elec I use the next day and the solar gets exported. If this happens a few days in a row, it kind of gets the idea and slowly ramps down the maximum charge. Then we have a day with more-or-less no solar and we're buying peak time electricity because it didn't charge enough overnight.

I'd like that slider to be there (as set out above) AND an API to be available on the LAN address so I can use weather forecasting software to predict how much solar I'm likely to generate the next day and automatically set the slider for me.
Does anyone from Tesla monitor these forums? Do these suggestions get picked up and at least considered I wonder?
 
Interesting, makes me wonder if the PW is right in our case. I think my house will never generate enough solar to even top up the battery in Summer so I want it fully charged during offpeak Octopus times. Is this impossible?

I charge my car every night but depends on the day (WFH or office day) the needs range from 5% top up to 50% top ups.
 
Interesting, makes me wonder if the PW is right in our case. I think my house will never generate enough solar to even top up the battery in Summer so I want it fully charged during offpeak Octopus times. Is this impossible?

I charge my car every night but depends on the day (WFH or office day) the needs range from 5% top up to 50% top ups.
A single empty Powerwall will charge at a max of 5KWh, so will take circa 2.5 hours to charge longer if it’s capped at 3.6kWh. 2PW’s can charge at twice that but usually at 7KWh (ish) they will fill in a four hour period . If your main fuse is rated more than 60 amps, you can charge a car at full whack too.
 
A single empty Powerwall will charge at a max of 5KWh, so will take circa 2.5 hours to charge longer if it’s capped at 3.6kWh. 2PW’s can charge at twice that but usually at 7KWh (ish) they will fill in a four hour period . If your main fuse is rated more than 60 amps, you can charge a car at full whack too.
Yes, but in an emergency, they will charge at 5kW each unless limited in their software.

When PW3 comes I will have mine limited to 3.68kW each; however, this really will restrict me as charging at 7.4kW will be pushing it, so I may turn off one Powerwall when charging the car.
 
  • Like
Reactions: Dilly
I've been fiddling with mine manually (adjusting RESERVE %age), and planning to build something with the API ... partly "once I have enough data" - which needs to include some seasonal data; but I have plenty of that now.

My PV will generate 15kW mid summer, and I don't think it is good to charge the PowerWalls full wack (they will charge a lot faster off PV than they do off grid), so if Sun is predicted to be good I charge one/both cars at same time as PowerWall to moderate the PowerWall charge rate, rather than fill the PowerWall full-bore than then use e.g. Zappi to detect export and redirect that to the car

Of course if I am diverting power whist PowerWall is charging I need a crystal ball to be sure that there will be enough sun to get my PowerWall to 100% just as the PV starts producing less than the house needs ... or clouds arrive. (I've got a reasonably good prediction model going using data available off-the-net)

Then I need to know what my needs are. Am I going anywhere? In which car? How much electricity does the house need tomorrow? - we have some specific activities like baking and washing that consume more juice on those days, but probably just saying High / Medium / "Away" would do. Along with "Must have XXX miles" in a specific vehicle "by tomorrow"

The other snag I'm trying to resolve is whether the car should charge at full 7kW always - i.e. potentially draw from the PowerWall - so that the car charges at optimum. Charging the car at 5AMPs is fine if the PowerWall is at 99% and the sun is about to go over the yardarm, but the efficiency down at 5AMPs is dreadful ... but I wonder what the optimum between Car and PowerWall is when PV is generating flat out - allowing for the occasional cloud passing over, or someone might turn the oven on for 30 minutes ... maybe constantly adjusting AMPs to match generation would be best, but I don't know what the response rate of the car is, or whether that is a good idea (efficiency-wise)

I'm also wondering what monitoring kit I should have in the house. There are a few things I could turn on to consume power. No point putting the immersion on as we have Solar Thermal ... but we have an Air Con that we use in one room on very hot days - clearly that could come on "on very sunny days", also pool filtration could run longer hours on sunny days, maybe pond filter too, perhaps a range of other things. But I'd need some monitoring for that ...

Let me know if you would be interested if I do get around to knocking up something on the API - something that factors in predicted solar generation, miles needed in the car, and whether house consumption is High / Medium / Away - I'd make more effort if I was in a collaboration! I do think I have some good ideas for optimisation that will minimise chance of grid usage.

By the by ... I've, recently, changed from setting RESERVE when I get up to, say, 10% (in case of power cut) and just letting the PowerWall discharge during the day; instead I now set it to the level that I expect to use after 6PM - and then reducing THAT to 10% at 6PM. It makes no difference to me if, on a cloudy day, my PowerWall runs out at noon, or much earlier and then discharges further after I drop the RESEVE at 6PM, and is then empty a bit later on. I'm going to pay for the same amount of peak-rate juice either way. And if-and-when we get incentive for using zero kW in the evening I'll be ready (although presumably the offer I will get will be based on my non-existent actual usage! but I'm going to be very happy challenging that :) )

Anyway, changing from discharging my PowerWalls fully during the day to keeping some for 6PM-onwards will, I hope, be saving a bit of peaker-plant gas ...

And, final point, for those of you "waking up in the middle of the night" to manually set a RESERVE. I have a run-rate of X% per hour in the late evening, so when I go to bed I set the reserve to current %age LESS the X% run rate x number of hours until Off Peak, so that basically as Off Peak starts, or shortly thereafter, the PowerWall will discharge and hit the RESERVE and stop discharging (i.e. use Grid). Then it needs to make up, from that level, to the desired end-of-off-peak level ready for "tomorrow" ... you'll still have to wake up to set that :) but at least the PowerWall won't have kept on discharging during Off Peak.

I presume it is possible to make that happen in API ... I just haven't got around to it yet.
 
Interesting experience over the past two days.

The day before yesterday, my PWs charged to around 90% overnight, and yesterday was wall-to-wall sunshine so there was no risk of running out of stored low-cost GO power. This is with thw reserve set to 20%. However, last night the PWs charged to 100% and as of now we are generating a measly 49W from 5.1kW net panels.

This tells me that Tesla AI is working on weather forecasts and not historical records, and it is clear that we should not unnecessarily manipulate the reserve % to buck the AI System.

In just over a week, my extra 12 395W panels come on-stream, so I'm not going to change any reserve % and instead see how the AI copes - I'm more hopeful than I was 6 months ago as the AI seems to be better.
 
I've been fiddling with mine manually (adjusting RESERVE %age), and planning to build something with the API ... partly "once I have enough data" - which needs to include some seasonal data; but I have plenty of that now.

My PV will generate 15kW mid summer, and I don't think it is good to charge the PowerWalls full wack (they will charge a lot faster off PV than they do off grid), so if Sun is predicted to be good I charge one/both cars at same time as PowerWall to moderate the PowerWall charge rate, rather than fill the PowerWall full-bore than then use e.g. Zappi to detect export and redirect that to the car

Of course if I am diverting power whist PowerWall is charging I need a crystal ball to be sure that there will be enough sun to get my PowerWall to 100% just as the PV starts producing less than the house needs ... or clouds arrive. (I've got a reasonably good prediction model going using data available off-the-net)

Then I need to know what my needs are. Am I going anywhere? In which car? How much electricity does the house need tomorrow? - we have some specific activities like baking and washing that consume more juice on those days, but probably just saying High / Medium / "Away" would do. Along with "Must have XXX miles" in a specific vehicle "by tomorrow"

The other snag I'm trying to resolve is whether the car should charge at full 7kW always - i.e. potentially draw from the PowerWall - so that the car charges at optimum. Charging the car at 5AMPs is fine if the PowerWall is at 99% and the sun is about to go over the yardarm, but the efficiency down at 5AMPs is dreadful ... but I wonder what the optimum between Car and PowerWall is when PV is generating flat out - allowing for the occasional cloud passing over, or someone might turn the oven on for 30 minutes ... maybe constantly adjusting AMPs to match generation would be best, but I don't know what the response rate of the car is, or whether that is a good idea (efficiency-wise)

I'm also wondering what monitoring kit I should have in the house. There are a few things I could turn on to consume power. No point putting the immersion on as we have Solar Thermal ... but we have an Air Con that we use in one room on very hot days - clearly that could come on "on very sunny days", also pool filtration could run longer hours on sunny days, maybe pond filter too, perhaps a range of other things. But I'd need some monitoring for that ...

Let me know if you would be interested if I do get around to knocking up something on the API - something that factors in predicted solar generation, miles needed in the car, and whether house consumption is High / Medium / Away - I'd make more effort if I was in a collaboration! I do think I have some good ideas for optimisation that will minimise chance of grid usage.

By the by ... I've, recently, changed from setting RESERVE when I get up to, say, 10% (in case of power cut) and just letting the PowerWall discharge during the day; instead I now set it to the level that I expect to use after 6PM - and then reducing THAT to 10% at 6PM. It makes no difference to me if, on a cloudy day, my PowerWall runs out at noon, or much earlier and then discharges further after I drop the RESEVE at 6PM, and is then empty a bit later on. I'm going to pay for the same amount of peak-rate juice either way. And if-and-when we get incentive for using zero kW in the evening I'll be ready (although presumably the offer I will get will be based on my non-existent actual usage! but I'm going to be very happy challenging that :) )

Anyway, changing from discharging my PowerWalls fully during the day to keeping some for 6PM-onwards will, I hope, be saving a bit of peaker-plant gas ...

And, final point, for those of you "waking up in the middle of the night" to manually set a RESERVE. I have a run-rate of X% per hour in the late evening, so when I go to bed I set the reserve to current %age LESS the X% run rate x number of hours until Off Peak, so that basically as Off Peak starts, or shortly thereafter, the PowerWall will discharge and hit the RESERVE and stop discharging (i.e. use Grid). Then it needs to make up, from that level, to the desired end-of-off-peak level ready for "tomorrow" ... you'll still have to wake up to set that :) but at least the PowerWall won't have kept on discharging during Off Peak.

I presume it is possible to make that happen in API ... I just haven't got around to it yet.
Thanks @WannabeOwner
Tinkering with the API is not something I want to be doing but I can understand that plenty will like the ability.
Charging a Powerwall should be as simple as charging a car. Tesla in their wisdom cannot even simplify that! The car has a percentage slider but no start AND stop times. The Powerwall has start AND stop times provided by TBC, but no percentage slider.
Both car and Powerwall try to be clever but fail in many instances. We the users have to rely on 3rd party kit and software to get a simple job done successfully.
Imagine how difficult it would be to charge a car to the required level if it chose for itself! A Powerwall is no different. Give me a user setting any day.
I have say that the current of controls can be likened to auto wipers and auto high beam and we know how good they are 😂
 
And, final point, for those of you "waking up in the middle of the night" to manually set a RESERVE. I have a run-rate of X% per hour in the late evening, so when I go to bed I set the reserve to current %age LESS the X% run rate x number of hours until Off Peak, so that basically as Off Peak starts, or shortly thereafter, the PowerWall will discharge and hit the RESERVE and stop discharging (i.e. use Grid). Then it needs to make up, from that level, to the desired end-of-off-peak level ready for "tomorrow" ... you'll still have to wake up to set that :) but at least the PowerWall won't have kept on discharging during Off Peak.

I presume it is possible to make that happen in API ... I just haven't got around to it yet.
So I already have a simple hacked-together Python script which, using cron on a MacOS server, sets and resets the overnight standby reserve at the appropriate times using the Tesla API. It's not the prettiest of scripts - I'm not a Python geek - but it's been working well here.

The tarball is here. There's a basic README.txt in the tarball explaining how to get it all set up, at least on a MacOS system. Linux or *BSD will probably be similar. There is no fancy web/gui front-end; all settings require editing a config file. Please don't ask for support!

For the last 6 months or so the "AI" has been fairly reliable so I enable the script only when I know our daytime use will be a fair bit higher than usual, but the forecast means the Powerwall will likely not charge sufficiently of its own accord at the off-peak rate.

Steve
 
Last edited:
So I already have a simple hacked-together Python script which, using cron on a MacOS server, sets and resets the overnight standby reserve at the appropriate times using the Tesla API. It's not the prettiest of scripts - I'm not a Python geek - but it's been well working here.

The tarball is here. There's a basic README.txt in the tarball explaining how to get it all set up, at least on a MacOS system. Linux or *BSD will probably be similar. There is no fancy web/gui front-end; all settings require editing a config file. Please don't ask for support!

For the last 6 months or so the "AI" has been fairly reliable so I enable the script only when I know our daytime use will be a fair bit higher than usual, but the forecast means the Powerwall will likely not charge sufficiently of its own accord at the off-peak rate.

Steve
Well done you. At least you get a get a night’s sleep! But wouldn’t it be nice if you could just set it to charge to x% during off-peak hours from the app 😁
 
Tinkering with the API is not something I want to be doing but I can understand that plenty will like the ability.

I'm a database programmer by trade, so hopefully that will help me!. I expected to find a 3rd party APP, similar to TeslaFi, that would allow doing all sorts of stuff with PowerWalls ... all I've found so far is "tinkering" :), and although I have the ability I lack the time to fiddle myself ... but maybe that's where I'm going to wind up.

I think there is benefit in having a number of users sharing a service, because the aggregated data could be used to improve the algorithm and allow some what-if calculations on historical data. For example, my use case is plenty of solar such that I cannot charge PowerWall and then, once full, charge the car ... I generate more than 7kW from PV, so I need to charge both PowerWall (more gently) and Car to not exceed the 7kW limit (I also don't think it is good to charge the PowerWall flat out - based on what I have read and given that Tesla charge it much more modestly during off-peak). Other people will have different scenarios and a necessity for one person might turn out to be handy for another.

I also like software to alert when "Something's wrong" - at work we spend a large portion of our timing investing in that because it dramatically reduces downtime on LIVE systems. For example: I think it is pointless, in this day and age, having a Thermostat that "calls for heat" and then, an hour later, when the room temperature has gone DOWN further! just carries on "calling for heat" rather than sending a "Something's wrong / Boiler Bust" email :) Aggregated data could be used to figure out 90:10 ranges and outliers could be reviewed / trigger an alert. My original PV trips once in a while ... first I know about that is when I go to do a manual reading once a week ... I think a "Prediction says its sunny, but you are generating 0kW" would be handy ...

There's a basic README.txt in the tarball

Awesome ... surprisingly clear for home-spun, thanks for taking the time to do that :)

Please don't ask for support!

I'm betting if I pontificate here some wild horses might come and find you :)
 
  • Like
Reactions: DenkiJidousha
Yesterday was a good sunny day and everything got topped up from solar.
knowing today and tomorrow would be pretty dull, at bedtime, the Powerwalls were at 77% so I set the reserve at 74% to fill them.
at 4.15 am they only filled to 78% and by 8am were at 72%.
So much for AI !