Recent Odd Charging Behaviour In Time Based Mode

[xWren]

In the last 2 weeks I have noticed odd charging behaviour when my Powerwall is in time based mode, compared to the previous behaviour 2021 through 2023. Any one else care to check their data, or ideas what is going on?

First some details of my use. While I have solar panels, my Powerwall usage includes charging from the grid during the low solar generation winter months. Tesla's prediction algorithms fail to manage Powerwall charging in the way I want. I have tried letting time based mode run the show, setting the prices etc., and it makes inefficient decisions (based on historic data) and costs me money. The coastal weather here is unpredictable as is our household use, so my local knowledge and intelligence are superior to any so called "AI". Hence I manage charging from the grid when I need it using scripts that adjust both mode and backup reserve percentage at the start and end of my 4 hour over night period of cheap electricity supply.

When I need more power than 4 hours at 1.7kW that self-powered gives I switch to time based for that period to get the higher 3.6kW rate (throttled in UK systems with official permission needed on installation to get the full 5kW).

Charging typically looks like this:

You see charging from the grid start at 12:30, ramp up quickly to 3.6kW and then drop off once the desired reserve percentage has been reached (it would stop at 4:30am if this was not done). My graphs consistently look like this for winters of 2021 through to end of December 2023. On colder nights (we rarely get below freezing) there is a gentle ramp up to the full rate which I assume is temperature related.

But in the last couple of weeks it does erratic things like this:

Yes in that case it did get to the 100% in the period, but what is it doing?

Other days It has failed to reach 100% having throttled the rate to a mere 1.7kW despite being in time based mode.

Clearly I had internet access to the cloud as charging from the grid started and stopped (so the backup reserve percentage was changed). The weather over night was no lower than 6C. The firmware has not changed since September 2022 (version 22.18.5).

What I would like most is self-powered mode to provide charging at 3.6kW, an improvement I was told by Tesla Support was coming back in November 2021. After all even their FAQ claims that 3.3kW is the optimum rate for the battery life. However having charging in time based behave erratically is a turn for the worse.

I can only assume that during "autonomous" time based mode the charge rate from moment to moment is managed from the Tesla Mothership, and that for 2024 they have changed the strategy so it no longer always uses the 3.6kW rate to get to the reserve percentage. This is bad news fore me - I only have 4 hours at cheap rate in which to charge my Powerwall.

@auxms75d how does this compare to your recent experience mentioned in the firmware version thread?

 

[BGbreeder]

Tesla's black box stance on algorithms and updates certainly does make planning difficult, and at times rather disruptive when things change with no apparent cause.

Hypothetical question; what if Tesla were to roll out an upper bound on the charging rate of 3kW? (Perhaps driven by real world data showing them that the higher rates of charge were detrimental in some way, so perhaps has rolled out said limit.)

What would be your preferred course of action? Would you purchase a second battery? Would it be a second Powerwall? Would you purchase a competing brand? Install more solar? Just live with the increased costs?

All the best,

BG

 

[xWren]

@BGbreeder it does seem unlikely that there is a battery reason (or even a grid supply one) to further cripple the charge rate in normal operating conditions. LiPo battery cells are normally charged at 1C, so on my Powerwall a 14kW rate would be fine (although everyone doing that would upset the grid supply hence the 3.6kW device limit in UK without permission). I think what they are doing with this erratic charging is either a bug the in Mothership software, and/or they are messing about with randomly spreading charging loads (which conflicts with my supplier offering a cheap off peak period that they want me to shift my usage into as much as I can, and a 3.6kW load is well within grid limits).
However hypothetically what would I do if the Powerwall was made even less locally controllable than it is now (for example ignores the backup reserve percentage)? I would seriously look to reverse engineering the firmware API so I could use the product that I own rather than be hampered by Tesla. Location means I can't have more solar, and no I would not buy more Tesla products, they are innovators but I do not like the flawed design assumptions that the internet is always there and they know best. I would also look at buying a GivEnergy battery based on a ROI analysis, if I was installing now that is definitely what I would buy rather than a Powerwall. But I admit ROI or not there is a limit to how much of Tesla giving customers no local control of devices that they own I will put up with.

For now I really want to know what is going on, is this a bug? Or if deliberate can Tesla be made aware of how this change in charging is negatively impacting owners like me that don't live in consistent Californian sunlight, and are using the Powerwall to shift energy consumption from the gird into off peak periods (in the way everyone wants us to).

 

[xWren]

Last night in time based mode my off peak charging was at 3.6kW, behaving just like the previous years. I'm increasingly convinced that those nights when it has not done that are some sort of bug.

The previous night was one of the erratic ones:

In all the erratic "spike" pattern charging night is almost like the Powerwall intermittently forgets that it is in time based mode (and can therefore charge at 3.6KW) and behaves as if it is in self-consumption mode which has very restrictive 1.7kW rate.

Does the Powerwall have to be continuously reminded by communication with the Tesla Mothership to use the time-based mode rate? So if for some reason communication is disrupted (by internet outage and the Gateway's 4G phone fallback not covering it, or Tesla server inaccessibility) then it defaults to a lowly 1.7kW rate. Does anyone know about this?

If only self-powered mode provided charging at 3.6kW as Tesla Support told me it was going to.

 

[JohnRatsey]

Last April I had a 2nd PW installed along the a backup gateway (not available when the 1st PW was installed in April 2018). Since then the winter charging in time-based mode has been at 5kW per PW until the charge level reaches 95% whereupon the charge rate reduces. Previously the single PW would charge at 3.5kW until 98% full before reducing the charge rate. That charge rate wasn't quite sufficient to fully charge an empty battery in the 4 hour Octopus Go cheap electricity period whereas the 5kW charging rate is higher than needed. I can't explain why your charging rate would reduce to 1.7kW before the battery charge level reached at least 95%. My 100A mains fuse should be good for 23kW so there's capacity for doing EV charging as well although that's not in my current planning.

When I had my system running in self-powered mode during the summer months it never charged from the mains but would receive up to 3.68kW from the inverter. I presume you can force mains charging by raising the backup reserve above the current charge level but I've never tried this.

As we've previously discussed elsewhere in this forum, my system takes a solar generation forecast into account when deciding how much off-peak charging to do when in time-based mode although it took several years of nagging Tesla before this happened. I wonder if you need to contact Tesla support to enable this feature. It's not perfect. It can handle forecast sunny days or cloudy days but struggles when the forecast says a mixture of sun and cloud which is also when a human would need to make an intelligent guess.

 

[BGbreeder]

Last night in time based mode my off peak charging was at 3.6kW, behaving just like the previous years. I'm increasingly convinced that those nights when it has not done that are some sort of bug.

The previous night was one of the erratic ones:
View attachment 1009704

In all the erratic "spike" pattern charging night is almost like the Powerwall intermittently forgets that it is in time based mode (and can therefore charge at 3.6KW) and behaves as if it is in self-consumption mode which has very restrictive 1.7kW rate.

Does the Powerwall have to be continuously reminded by communication with the Tesla Mothership to use the time-based mode rate? So if for some reason communication is disrupted (by internet outage and the Gateway's 4G phone fallback not covering it, or Tesla server inaccessibility) then it defaults to a lowly 1.7kW rate. Does anyone know about this?

If only self-powered mode provided charging at 3.6kW as Tesla Support told me it was going to.

I would try to lean on Tesla to up your charge rate to 5kW, if you can get your local provider to sign off on the higher charge rate.

You raise an interesting point about connectivity. Is there a possibility that there is a local to your house WiFi issue, or to the internet in general, during the overnight hours? Have you checked your voltages overnight during the charging period with a voltage/power logger, in case there is some other issue?

All the best,

BG

 

[auxms75d]

@auxms75d how does this compare to your recent experience mentioned in the firmware version thread?

I'm not sure my data/experience would apply as i've got several scripts that monitor and control both my PW charge as well as the car charging and pre-conditioning. This was originally required because the max panel rate was set incorrectly by tesla during installation and it took them several months to figure this out and update the value. When installed they set it such that anytime the house pulled more than 18kwh the pw will go into standby mode. To ensure the PWs could grid charge I needed my scripts to schedule (and limit) charging of the evs while being aware of the pw charging status. So my system changes the reserve set point to 100% nightly to fully grid charge, and then return the set point to 20% once fully charged. Once the PW was fully charged I could let the cars charge at their normal rates and it wouldn't matter that the house was consuming >18kwh.

With that setup before the FW upgrade I would see the PWs charge at a max rate of 5kwh/PW with a ramp down in charge rate when the PW reached >98% SOC; when the ambient air temp was below 10c the PWs would charge at <1kwh/PW for about 10-15 mins and then jump to 5kwh. I believe the thermal protection would heat the batteries before beginning the max charge rate.

 

[xWren]

@JohnRatsey interesting that a 2PW setup gets 5kW, did you get permission from your DSO for that? Maybe your installer simply enabled it (good for you if so), someone I didn't go with offered to do that for me so it must happen. Yes you can force mains charging by raising the backup reserve above the current charge level, and in self-powered mode the rate used in 1.7kW.

@BGbreeder yes I have wondered if it was worth applying for permission for 5kW, it is likely that the admin fee has come down. I wonder if I'll have to pay someone with the installer app to come out and configure the Gateway too or would Tesla support do it remotely? Then again since I mostly use self-powered mode the real issue is getting even 3.6kW in that mode.

Flawed algorithm interaction with my internet connectivity is the main suspect. I did have an internet problem and swapped to a much slower 4G based connection on one of the erratic nights, and maybe there have been other brief outages. But shouldn't the SIM in the Gateway cover for lack of internet at least when it comes to control (if not full data collection)? Does "autonomous" mode actually mean highly internet connection dependant? On Jan 16th for example my internet was all good best I can tell, so not sure what to make of it. There are definitely times in the past when Tesla servers are unavailable , I know that because my scripts have failed with comms errors (I have them re-try), it just isn't a 24/7 system. If only charging in self-powered mode was 3.6kW, and control of the PW was local.

@auxms75d I really appreciate your script usage to get sufficient control of your system. The pre FW upgrade behaviour you describe matches my normal experience too. Could it be that you had some internet connectivity issues or other server access problem since the FW change?

I guess it is up to me to experiment: stay up until the early hours and see what interrupting my internet connection does to the time based charging behaviour.

 

[BGbreeder]

@xWren Please forgive the question, but you do have have a fixed IP address reserved for your gateway in your router, right?

Is your gateway on WiFi, or wired via Ethernet? If it is the former, you might try an overnight test with a cable back to your router directly to see if anything changes.

You might also consider running several parallel, continuous, pings overnight to your gateway and to several outside locations to probe the quality of your internet connection, just to rule that out as a potential factor.

All the best,

BG

 

[xWren]

@BGbreeder thank you for your interest, questions welcome. No I am unable to fix the IP address in my router. I am rural, no fibre here and 3 miles of overhead copper wire can not deliver stability nor speed, so I have internet via radio and as part of that I am not granted access to the router that allocates addresses. I use fixed IP addresses for every device that has a UI that allows me to set it at the device e.g. printer and NAS drives, but the gateway does not offer that. That said, the IP address allocated to it has been the same for as long as I can remember, and certainly has not changed this month.

The gateway is on WiFi, it is located outside and running a LAN cable to it would be tricky (across a passageway and via an open window, not something I want to do over winter).

The inverter and clamps on the solar are inside the house, we have Neuro meter connecting to the gateway via WiFi too (because of cable routes) providing the solar generation data. There was a period of firmware that had all kinds of connectivity issues with the Neuro, but with v22.18.5 has been pretty stable hence I am reluctant to ask for an update.

Could charge rate from grid in time based mode also be sensitive to the connection to the Neuro meter measuring solar (even when there is none because it is night)?

My persistent thought is that it is a design weakness to have made time based mode so dependant on continuous communication with the Mothership (I'm still betting that is what I'm seeing, yet to experiment, regardless of where the break in comms happens), and a total flaw to needlessly restrict self-consumption mode to charging from the grid to a meager 1.7kW. Good design would be that the PW charges from grid at a permitted rate for the installation and current battery temperature whenever it falls below the backup reserve percentage regardless of the mode.

 

[swedge]

Pure guess here: Is this possibly a temperature issue? Perhaps PW does not like to charge quickly when it is cold, but charging warms it up so after a while of slow charging it can speed up?

The PW2 spec sheet (link) has a footnote which says:

2In Backup mode, grid charge power is limited to 3.3 kW.​

Otherwise, the charge rate max is 5 kW, and I think I have seen it grid charge at 5 kW when Storm Watch is active.

At least here in the US, the User page in the GateWay lan web access Network menu allows setting a statice IP address.

 

[JohnRatsey]

@JohnRatsey interesting that a 2PW setup gets 5kW, did you get permission from your DSO for that? Maybe your installer simply enabled it (good for you if so), someone I didn't go with offered to do that for me so it must happen. Yes you can force mains charging by raising the backup reserve above the current charge level, and in self-powered mode the rate used in 1.7kW.

The installers said that's the way they do the configuration although there was correspondence between the installer and the DNO prior to installation. The grid code used is 50Hz_230V_1_G99_2018_UK. I note that the National Grid G99 guidance document differentiates between generation capacity and export to the grid which I don't recall seeing for the previous system. Anyway, that aspect irrelevant to your problem of charging rate and I see nothing in that document about charging rates. I wonder if the battery assumes that the maximum charging rate is the same as the maximum generating capacity.

The G99 code became applicable for installations after April 2019 although I don't know how long elapsed before Tesla added provision for that code to the gateway software. IIRC my first PW was configured to G83 which applied the 16A (3.68kW) limit to both generation and charging. Perhaps someone could change your grid setting to G99 in the PW setup? I found the old 16A generation limit to be a nuisance as it was easy for the domestic load to exceed that and draw some power from the grid. 10kW generating capability (5kW per battery) means my system will be better able to run a heat pump as well as other equipment.

 

[BGbreeder]

@BGbreeder thank you for your interest, questions welcome. No I am unable to fix the IP address in my router. I am rural, no fibre here and 3 miles of overhead copper wire can not deliver stability nor speed, so I have internet via radio and as part of that I am not granted access to the router that allocates addresses. I use fixed IP addresses for every device that has a UI that allows me to set it at the device e.g. printer and NAS drives, but the gateway does not offer that. That said, the IP address allocated to it has been the same for as long as I can remember, and certainly has not changed this month.

The gateway is on WiFi, it is located outside and running a LAN cable to it would be tricky (across a passageway and via an open window, not something I want to do over winter).

The inverter and clamps on the solar are inside the house, we have Neuro meter connecting to the gateway via WiFi too (because of cable routes) providing the solar generation data. There was a period of firmware that had all kinds of connectivity issues with the Neuro, but with v22.18.5 has been pretty stable hence I am reluctant to ask for an update.

Could charge rate from grid in time based mode also be sensitive to the connection to the Neuro meter measuring solar (even when there is none because it is night)?

My persistent thought is that it is a design weakness to have made time based mode so dependant on continuous communication with the Mothership (I'm still betting that is what I'm seeing, yet to experiment, regardless of where the break in comms happens), and a total flaw to needlessly restrict self-consumption mode to charging from the grid to a meager 1.7kW. Good design would be that the PW charges from grid at a permitted rate for the installation and current battery temperature whenever it falls below the backup reserve percentage regardless of the mode.

Vis-à-vis the gateway address; my intent was for the gateway to have a locally reserved IP on your local network, e.g. 192.168.XX.YY. I believe that the Tesla Gateway address can be(is) specified in the configuration menu, but if the address isn't reserved, your local router can create conflict and poor connectivity just by handing out that address to another device via the DHCP address conflicts. The neurio is said to also benefits from a reserved local address, although as far as I can tell, the gateway to neurio communication is strictly MAC address to MAC address. (Which are both also specified in the configuration menu, and can lead to issues if the IP address doesn't match the MAC address. BTDT.

Internet via radio (microwave?) can certainly be subject to extra interference, but extra interference specifically midnight to 04:00 would seem unusual in a rural area, at least to me.

Is the Powerwall discharging in the 20:00-24:00 time period? If so, it is unlikely to be below 20F.

I suspect that your analysis is the correct one. Have you logged into the configuration menu to see what your various site limits are?

All the best,

BG

 

[xWren]

@swedge thank you but no it is not a temperature issue. Last night was our coldest this winter (we got to -2C or 28F), and the charging in time based mode was exactly as I would expect with some ramping up at the beginning because of warm up, but no flats at 1.7kW and then odd spikes of 3.6kW :

It just got on with charging up to the backup reserve percentage at 3.6kW as soon as the battery was warm enough.

The PW2 spec sheet (link) has a footnote which says:

2In Backup mode, grid charge power is limited to 3.3 kW.​

Otherwise, the charge rate max is 5 kW, and I think I have seen it grid charge at 5 kW when Storm Watch is active.

Interesting, do you have a link for that @swedge , the spec I have doesn't say that but I would love to have something to approach Tesla support with.

At least here in the US, the User page in the GateWay lan web access Network menu allows setting a statice IP address.

I'll take another look. I don't think the IP address is changing and causing an issue but always happier to be able to fix it.

 

[xWren]

Vis-à-vis the gateway address; my intent was for the gateway to have a locally reserved IP on your local network, e.g. 192.168.XX.YY. I believe that the Tesla Gateway address can be(is) specified in the configuration menu, but if the address isn't reserved, your local router can create conflict and poor connectivity just by handing out that address to another device via the DHCP address conflicts. The neurio is said to also benefits from a reserved local address, although as far as I can tell, the gateway to neurio communication is strictly MAC address to MAC address. (Which are both also specified in the configuration menu, and can lead to issues if the IP address doesn't match the MAC address. BTDT.

Internet via radio (microwave?) can certainly be subject to extra interference, but extra interference specifically midnight to 04:00 would seem unusual in a rural area, at least to me.

Is the Powerwall discharging in the 20:00-24:00 time period? If so, it is unlikely to be below 20F.

I suspect that your analysis is the correct one. Have you logged into the configuration menu to see what your various site limits are?

@BGbreeder I understood what you were suggesting for a reserved IP address and the problems when not. The addresses on my local network are managed by a router (in my house) that I do not have access to configure. I know, I had thoughts about that, but this is all part of my internet via radio system, that runs in a 5.8GHz range and is encoded and pretty free of interference at my location. We have only had one problem day with interference caused by equipment on a Royal Navy ship that was moored off-shore (the forces do not have to comply with civil radio regulations). That made for a different kind of support conversation! When our internet goes out is is usually hardware failure of some piece of radio kit in the chain and someone has to get up a mast and replace it.

It is pretty rare for my location to get down to 20F, one of the benefits of being coastal and the warm flow of the Gulf stream.

 

[BGbreeder]

@BGbreeder I understood what you were suggesting for a reserved IP address and the problems when not. The addresses on my local network are managed by a router (in my house) that I do not have access to configure. I know, I had thoughts about that, but this is all part of my internet via radio system, that runs in a 5.8GHz range and is encoded and pretty free of interference at my location. We have only had one problem day with interference caused by equipment on a Royal Navy ship that was moored off-shore (the forces do not have to comply with civil radio regulations). That made for a different kind of support conversation! When our internet goes out is is usually hardware failure of some piece of radio kit in the chain and someone has to get up a mast and replace it.

It is pretty rare for my location to get down to 20F, one of the benefits of being coastal and the warm flow of the Gulf stream.

Thanks! I was just checking to ensure that we were all on the same page.

I have some experience with the 5.8GHz relays, and it takes quite a bit of rain to knock them off their game. I could see that a RNS offshore could generate some issues.

I don't have any other useless suggestions, beyond asking your installer to turn up the allowed import rate, although I do understand that for much of the rural UK, it isn't a trivial request.

All the best,

BG

P.S. I don't see any charge current specification in the UK (GB) Powerwall specification.

https://www.tesla.com/sites/default/files/pdfs/powerwall/Backup_Gateway_2_Datasheet_en_GB.pdf

 

[xWren]

Yes @BGbreeder the 5.8GHz signal is pretty robust, weather has never made a noticeable difference (well except for when high winds and rain physically took out the mast mounted equipment). The prior experience with internet via overhead copper wire was that instability and noise rose (and thus speed dropped dramatically down to 4Gb and below) whenever the wind blew from the 'wrong' direction. Both stretching of wire joints and factures, and cable movement within the range of emf created by adjacent power supply cables contributed to a very unhappy consumer experience. Baring Navy ships radio is cool. Happy to give you Californians a vision of life in other parts of the world, even if I'm not average for the UK either.

Shame that there are no promises in that UK (GB) Powerwall specification about self-powered mode charge rates, and backup mode no longer exists as I understand it (removed from the API anyway).

Thinking more about charging rate, I'm actually happy with 3.6kW on our 60A fused transformer supply, leaves headroom for other uses in that off peak 4 hour cheap period (like charging the EV, or running the dishwasher over night too) while the battery charges. The power grid at our location could cope with 5kW no problem, it is just an admin/paperwork thing for our DSO. I just wish that I could get 3.6kW import on self-powered mode, and could set the backup reserve percentage locally of course.

Thank you so much for your input and interesting discussion.

 

[BGbreeder]

@xWren Always nice to learn more about other parts of the world. (Though I have spent a fair bit of time up "North", and; have some experience with rural UK life... It makes me smile every time I drive up the M1 and see the roadside sign that says "The North".)

I do wonder how the grid in the UK will evolve as solar and EVs expand. 60A single phase gets used up rather quickly with an EV, an electric boiler, and a heat pump... It will be interesting; I am sure that more than a few rural homes are still on fuse wire.

All the best,

BG

 

[swedge]

Interesting, do you have a link for that @swedge , the spec I have doesn't say that but I would love to have something to approach Tesla support with.

Sorry, I thought I copy/pasted in link.

https://www.tesla.com/sites/default/files/pdfs/powerwall/Powerwall%202_AC_Datasheet_en_northamerica.pdf

 

[auxms75d]

Oddly enough my system appears to now be following TBC and actually back to charging at 5kwh/pw; something seems to have changed exactly at 1am local time from 12:03am to 1am it was charging at ~1.5kwh/pw and suddenly it reverted back to charging at full speed. I'm waiting for my tarrif to change and see if its gone back to the expected TBC algorithm.