Tesla: we need more control over our PowerWalls

[mw963]

With regard to the Time Based Control, I am noticing that - since the system has been updated to allow one to put in the cost per kWh of each period of a tariff - my night time charging is behaving more sensibly. For example a few days ago we had two days on the trot of constant sunshine, with about 20 kWh of solar generated during the day. In spite of this I was pleased to see that the battery still charged up to 100% on the following nights, which given the dull days that followed the sunshine was exactly the right decision for this time of year. Prior to this change a couple of days of sunshine would have meant that the battery only partially charged on the following nights.....

I'll keep an eye on it, but just maybe things are improving a bit....?

 

[ST63]

I'm late to the party here - found the thread after seeing some odd behaviour on my PW2. We had the system installed Nov last year and are on Octopus Go. Until the last few days the system has performed as expected on TOU - charging to between 90-100% every night off peak, mostly to 100% @3.5kw charge rate. Our solar production has been Dec: 2.2 Avg, 6.5 High, 0.4 Low, Jan: 3.5 Avg, 7.5 High, 0.7 Low, Feb: 6.5 Avg, 15.6 High, 0.9 Low. Now the last 5 days of Feb were 10.6, 13.7, 15.0, 15.6, 3.1 . Monday , Tuesday, Wednesday this week the PW has only charged to 50% off peak. It clearly thought this week was going to be the same as the 4 good days. Frustrating. I agree with other comments - a simple setting 'battery % to achieve in off peak' would solve this. I might try the Rasp Pi but think its probably a bit beyond me....a simpler solution would be very welcome (and not getting up at 00:30 to change the setting !

 

[mw963]

@ ST63 - how odd that your and my system are doing the opposite. Like you I'm not sure I'm up to the Raspberry Pi solution, I bought the kit but then gave up when other things got in the way.

Have you put in £ values for your tariff on the app? I'm wondering whether the fact I put in a high rate (31 p per kWh) for the peak rate has made the app realise that I really REALLY want it charging up as much as possible when it can during the night....?

Curiously enough I've just put the phone down to Tesla. I rang to complain that one of their staff has ignored four emails over the last six weeks from me requesting a response from him. Unfortunately he wasn't in the office today, but another member of staff has promised to pass on my complaint. In addition I asked for advice given that for me the app is STILL running so slowly as to be awful. Unfortunately on that front they could see nothing wrong, but I stressed to them that the unofficial Powerwall app for Android loads the data in under 10 seconds, whereas the official Tesla app takes about a minute to update for me.

 

[ST63]

@mw963 - I had the cost set to my actual Go peak rate which is still pretty low (not for much longer I fear) but I've just now banged that up to 30p to see what it does.... have no problem with the iOS app running on a iPhone 2020SE.... I'll see how it goes

 

[mw963]

Thanks ST63. Keep us posted.

In view of what's going on I ended my Octopus Go Faster (5 hour 0130 - 0630, 5.5 p / 16.25 p per kWh) six months early, switching to the "new" version of exactly the same tariff (same times, 8.25 / 31 .25 p per kWh) but fixed for a year. I checked with Octopus that this was "allowed", my thinking being that it bumps up my price early but fixes it throughout next winter, when things of course may have improved, but just as likely will have got even worse. As we're using next to no peak rate electricity thanks to the battery it seemed more important to hang on to a reasonably cheap "night" rate, even if it means paying a little more earlier. In other words I hope I've beaten TWO price rises, the one coming in April, and another one due in October.

Hopefully the sun will cover us for most of our consumption fairly soon - like you I only installed last autumn so I've yet to enjoy the summer conditions.

 

[mw963]

Sadly we're back to the night time charging making poor decisions - I trusted ours last night when it started charging on TBC at 0130, but should have checked later in the night when I woke up, as it stopped at 80%. Now we have a cold morning, a dull day forecast, and a battery that isn't full.

I must try and get that Raspberry Pi working. I've completely lost faith in the Tesla app and algorithm.

 

[ST63]

Yes, fiddling with the peak/off peak rates doesn't appear to influence charging behaviour. I believe it only looks at generation history - last 3 days - and also typical household consumption. No attempt to look forward at all.....

 

[JCUK]

ST63 I observed this a few years back, the powerwall forecasts solar generation on the best figures in the past 3 days. If you want to force it to charge overnight you will have to set the peak sell rate considerably higher than the off peak cost rate, but a bit lower than the peak cost rate. When the off peak periods ends you will need to change the peak sell rate back to zero, or it could send to grid.

 

[aesculus]

the powerwall forecasts solar generation on the best figures in the past 3 days.

How did you determine this was the algo?

 

[JCUK]

I have had the powerwall since June 2017. When they eventually provided the app I was informed that the powerwall did not need manual controls as it used local weather forecasts to predict solar generation. When it became obvious that this was not true I kept a daily record when not on self powered mode. It was easy to spot that it only charged to 100% in off peak if there had not been any significant solar generation in the previous three days. I have an 8Kw system, and there can be large differences between sunny days and overcast days.

 

[aesculus]

The only algorithm I have been able to detect in Cost Savings Mode is near the weekend (for me a rate change) where if it feels, based on recent experience, that it won't need as much energy, or thinks it can generate more with less use, it will continue to discharge in partial peak rather than just peak.

This works OK in California during the summer but with last years fires there were many days our solar production cratered and I found the PWs discharging when they shouldn't have, so I manually had to bump up the reserve during those periods to prevent this.

 

[JohnRatsey]

I'm wondering if Tesla is finally looking at the UK weather forecasts. I was surprised when the battery fully charged itself early yesterday as if it had seen the weather forecast of solid cloud and rain. The previous behaviour would have considered that with some excess solar generation had gone to the grid on the previous three days then there would be no need to fully charge. Then, last night the battery didn't do any off-peak charging which was a sensible decision given the sunshine forecast for today. This more intelligent behaviour must have been introduced very recently as my Powerwall was also caught out by the grotty weather at the start of March following the sunshine in late February.

However, it will only be good luck if the battery manages optimum behaviour on those days when the forecast is a mixture of sun, cloud and rain but it's a big step in the right direction if the algorithm is now aware of the bigger changes in the weather.

 

[mw963]

Whilst I still don't trust my TBC an inch, and still regularly wake up anyway during the night and adjust it, as I said on 2nd March (upthread) I do think that it's making some rather better informed decisions, and like you JohnR I'm finding that it does seem to have moved away from the "it was sunny yesterday and today so I won't bother charging up the battery as it's bound to be sunny tomorrow".

If it's the case that they have done something to the algorithm why on earth not contact those of us who complain regularly, and have the humility to ask "is it working better?". On that subject I STILL cannot get one particular individual at Tesla Europe to reply to me, in spite of countless emails and a phone message left for him. I shall try once again, and then name and shame him here, I'm fed up with the arrogance of Tesla and its European Service Dept.

 

[xWren]

@JohnRatsey @mw963 I'm afraid I think you are fooling yourself if you believe the algorithm has changed. We happen to have had a very stable period of weather in the UK, I notice that I have not been making much adjustment via my scripts. Once the random weather returns so will the bad TBC charging decisions. Then again summer, with better solar generation and less load shifting etc., the algorithm mistakes have much less impact.

Of course someone could always try asking Tesla support if they now use local weather forecasts (not storm warnings for storm watch, but normal weather, cloud cover etc.) to determine how much to manage the Powerwall.

 

[JohnRatsey]

@JohnRatsey @mw963 I'm afraid I think you are fooling yourself if you believe the algorithm has changed. We happen to have had a very stable period of weather in the UK, I notice that I have not been making much adjustment via my scripts. Once the random weather returns so will the bad TBC charging decisions. Then again summer, with better solar generation and less load shifting etc., the algorithm mistakes have much less impact.

The weather has been stable except for Wednesday when it rained all day and my panels generated less than 1kWh. Past behaviour has shown that the battery would not have fully charged itself before that day and would have charged itself afterwards. This time it anticipated the weather and behaved sensibly. See SN36AJ 4.050kW (select right-most blob under the date selector to see the battery data. 8th May 2021 provides an example of the old behaviour. Or you can rummage through my solar generation graphs and choose another example.

There will be plenty of occasions, however, when interpretation of a sunny/cloudy/rainy weather forecast can get the estimated solar generation wrong. At one stage I tried doing my own mental estimates of solar generation based on the weather forecasts and concluded that it's difficult to get it close unless there's solid sunshine or solid cloud. One input to Tesla's algorithm must be the maximum solar generation for the site for that time of year and sensibly this is based on actual data.

 

[mw963]

Thanks JohnRatsey, and for access to the graphs. I'm now going to ask what I suspect is a stupid question - if it's the Tesla algorithm controlling your system, how does Solar Edge come into it. What am I missing?

I have access to a neighbour's monitoring system via Solar Edge - their solar and battery system ISN'T anything to do with Tesla, and I've always wished that our interface (ie the Tesla App) was as quick and clear as theirs using Solar Edge. So is it the case that I could somehow get SolarEdge to monitor my system, in addition to the truly awful Tesla App? I'd always assumed that our neighbours used SolarEdge because that was *their* system controller, and that we were lumbered with the Tesla App because that was *our* controlling system.

Sorry if I'm being dumb....?

 

[JohnRatsey]

Thanks JohnRatsey, and for access to the graphs. I'm now going to ask what I suspect is a stupid question - if it's the Tesla algorithm controlling your system, how does Solar Edge come into it. What am I missing?

I provided a link to the SolarEdge data so you can directly see the changes in solar generation and particularly a big change after several days of fairly consistent output. You can then look up that day at PVOutput and see what happened before and after. The next test will be to see what happens when the current spell of sunny weather ends.

 

[mw963]

Thanks - but how do *I* get Solar Edge to monitor MY system in the same way that YOUR Tesla system is being monitored by Solar Edge?

Or am I still missing something?

 

[Gwgan]

last night the battery didn't do any off-peak charging

According to US support docs, and my own experience, “When Powerwall is installed with solar, it will not be able to charge from the grid. “ except when in Storm Watch mode. Are things different in the UK?

 

[mw963]

@Gwgan - yes, very different, it's a real money saver being able to charge up at night on cheap rates during the winter and then use that energy during the day instead of drawing from the grid.

My understanding is that you over in America - and I may need correcting on this - get some sort of subsidy on a lot of installations, whereas here in Britain any subsidies are pretty much a thing of the past. I think it's the case that over in America it would be felt to be unfair to subsidise installations and then also allow them to become money spinners via "load shifting".

But my knowledge isn't that great so am more than happy to have someone explain it properly!