-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Powerwall 2: Technical
- Thread starter Aussie
- Start date
-
- Tags
- Powerwall Tesla Energy
PlugNPlay85
Member
Hey,
Anyone having issues controlling your Powerwall 2 using the Tesla App? With the PW at 0% charge, when I select backup mode the PW doesn't start charging immediately. The PW should start charging from either solar or grid or both correct?
If I switch to self powered mode at night and set the reserve to let's say 5% the charging is also delayed like 10 - 20min. PW will start charging from the grid but will charge past the 5%. I thought that the PW will only use the grid up to the reserve limit and continue charging from the solar source once available?
Anyone having issues controlling your Powerwall 2 using the Tesla App? With the PW at 0% charge, when I select backup mode the PW doesn't start charging immediately. The PW should start charging from either solar or grid or both correct?
If I switch to self powered mode at night and set the reserve to let's say 5% the charging is also delayed like 10 - 20min. PW will start charging from the grid but will charge past the 5%. I thought that the PW will only use the grid up to the reserve limit and continue charging from the solar source once available?
Ulmo
Active Member
- I've noticed delay between updating the Custom settings and the reaction by the system. Last time I tested, I waited 15 minutes, and it didn't comply; I woke up the next day and noted that it had complied within about 30-45 minutes.
- That seems like a proper way to operate: recharge from grid to get to backup minimum and don't charge from grid to go past backup minimum. I wonder why it's not working that way.
- I think an AI is the best way to do this, but then I think of all the stupid people I know, and I realize it better be a smart purpose-educated AI.
Bigtanuki
Member
Anyone out there have any experience with the Powerwall 2 under brown out conditions? About 75% of our "outages" are brown outs with voltage dropping to ~60 VAC. Led bulbs stay lit (dimly) but I try to kill all my motor loads that could be damaged like the refrigerator and freezer. Any help would be appreciated.
This has been my experience too and it was explained to me by the installer that it would take "about 15 minutes" for changes in the app to be pushed to/register with/cause the gateway and battery to change how they're routing energy. I'm curious as to why the delay exists though.
PlugNPlay85
Member
So I talked to Tesla support and the reason for the delay is the server. After changes are made in the app, it gets sent to a server. The server refreshes every 30 minutes to an hour. He made it clear to say that the refresh time doesn't start when changes are made. So you may get lucking and have changes made immediately or it could take as long as an hour to see any changes.
Fine, I'll be the nosy one to ask. What's happening between midnight and noon 7/29 and 7/30? Rapidly cycling AC?Graph of my power flow:
(Above dynamic; URL may not work forever.) Static file as of now:
Ulmo
Active Member
Haha, I don't think anybody had to be nosy. I posted it; you can pick it apart for all I care. That's my electric space heater. Every time it clicks on, it takes a set amount of power; then it shuts off when it gets warm enough. It augments my now working again small GPU cryptocurrency miner; it isn't earning me much -- it already cost me more money to repair due to maintenance than it earned me (by a little) (a disk drive failed) -- maybe it will end up having saved me problems later since it shook out the bad parts from that computer. I had to invest in some better insulation and door seals to get the use this low during the winter. It's (almost) always cold in Aptos at night.
As long as you're snooping, I came make a game out of this: guess when I get up for and leave for work tomorrow (after you see the graphs). Of course, it would help to see the finer resolution ones available at http://104.131.152.195/
At some point, I might do what @wk057 already did and retract the house and powerwall use information; that's more private. But for now, I want to leave it up so everyone understands how it interrelates with the PowerWall installation.
Last edited:
Ulmo
Active Member
Ulmo
Active Member
If you physically catch it in time to save a reserve after backup use has almost depleted the batteries (i.e., a four dimensional two line intersection at one point, a rough thing to do, basically meaning you have to be monitoring it during a power failure and be home), you can physically turn off all or some subset of the PowerWalls, either at their on/off switch on the batteries, or the breaker where they feed into the breaker panel. I presume this would be enough to save that energy until a time that you can turn them back on (presumably during enough solar input), thus re-starting up your backed up microgrid, enough to get the solar system to think that it's safe to start charging. At that point, the batteries will start to charge from the solar power, and you can replete them.Also, I haven't looked in awhile, so I need to go through the big posts that happened recently, but one thing I've thought a bit about is what happens in an extended outage when you exhaust your battery. You get a double-whammy at that point:
For me, the solar panels and Powerwalls are on a separate breaker panel that feeds into the backup gateway separately from the main breaker panel. I don't know anything about the details of the system architecture, but I could dream up something silly like the backup gateway isolating the main breaker panel from the panels and batteries and leaving just the panels powered until the sun rose and they produced enough power to charge the batteries above a certain reserve point. That might be the most sensible way to go. Unfortunately, to test this theory, I'd need to produce my own extended outage and I'm pretty sure I don't want to.
- Your solar panels need to see AC voltage before they'll generate, but if your battery is out, then you can't get AC voltage out of it.
- If the system is jump-started somehow (e.g. something tells the batteries that solar generation is feasible and they dump some reserve battery capacity in order to get the panels going again, or you plug in a generator, or something), then your panels need to generate enough to run whatever'll turn on in your house at that moment. Otherwise, you get a brown-out and you're back to the same problem.
The trick would be to always physically turn them off before it's time for them to be depleted again. That means black out for your home, every time.
I have not tested this. Basically, it could be tested easily:
0. This is a nice time to turn off everything in your house. (This is a blackout test. Also, you could probably do brownout tests, too, as you mentioned. Shutting down power to something while it has a load on it tends to damage switches with contactor metal melted splatter from the in-use spark across the gap during disconnection or connection when the metals of the two sides of the circuit are close enough for electricity to continue passing, especially on components that have not been made to handle this situation on a regular basis. For this reason, I would not do any brown-out tests myself.)
1. Turn utility grid connection off.
2. Turn solar input off.
3. Turn batteries off.
4. Confirm no power.
1. Turn utility grid connection off.
2. Turn solar input off.
3. Turn batteries off.
4. Confirm no power.
Now, the test is if the batteries can restart the solar. So:
5. Turn on batteries.
6. Confirm power if you like. You might need to turn on some devices to give the PowerWalls something to do and generate real action.
6. Confirm power if you like. You might need to turn on some devices to give the PowerWalls something to do and generate real action.
Indeed, after this test (steps 0-10), it might even be interesting to try this all over again with zero load at all to see if the solar can detect power without any load at all on the batteries, and then to see if they start to charge the batteries; that's an edge condition that could have different results than a loaded system.
7. Turn solar input on.
8. Confirm that the solar will now start to input into your system. This can be done by proxy by skipping to step 9.
9. Confirm that the solar charges the batteries.
7. Turn solar input on.
8. Confirm that the solar will now start to input into your system. This can be done by proxy by skipping to step 9.
9. Confirm that the solar charges the batteries.
You could have a complication that your Internet takes half an hour to an hour to restart, because you just turned everything off; in that case, one way to hurry up the process is to know the LAN IP address of your PowerWall gateway switch (not to be confused with ethernet switch or Internet/wifi/LAN gateway!*) after you turn on your LAN DHCP server (e.g., probably your little Internet closet full of stuff -- but usually you would need to know which one is the DHCP server so you can query it for the device list, or you can just try http://a.b.c.d/api/meters/aggregates for each IP# in your LAN range (check your computer's LAN IP# to get a guess)) (the PowerWall switch might have a new LAN IP# from DHCP after restart), and go to its address to get its current status, like http://192.0.0.5/api/meters/aggregates and see what it gives back to you. You're looking for battery: instant_power: <number_in_watts>; if it's negative, then it's charging, and if it's positive, it is discharging.
Then you know it works with no utility connection.
10. Turn on utility connection.
Family is here, so I cannot do step 0.
* Yes, the Tesla PowerWall Gateway is a switch, and it is called the Gateway or the switch or the controller. (It's a switch and power gateway between the various power components.) Very easy to get confused when discussing data networks too at the same time, since both power networks and data networks have gateways and switches, naturally; both are their own types of networks and the Tesla gateway is on both kinds.
Last edited:
Ulmo
Active Member
Confirmed: the Gateway starts using the smartphone app's new Customize settings (mode setting of Backup-only mode or Self-powered mode, and % Reserved for Self-powered mode) almost exactly on the half hour (18 seconds past, so at first I thought it was using GPS time (see GPS, UTC, and TAI Clocks for the 18 second offset), but then it would have been 18 seconds early, not 18 seconds late). I have family over which go through more in a day than my solar can provide, so I got smart and set the % Reserved higher than the % Charged at 11PM (I should have done it any time between 10:31PM and 10:59PM is what this info means). Here's a few captures of the day's charging as a result that demonstrate it:
This zoom in to the 11:30PM transition shows the timing better:
This zoom in to the last hour also shows the later timing better:
I'm having trouble believing the 11:30PM crossover skips so many datapoints. Zoomed more in plotted as points:
As you can see, my sample rate is every half a second, yet the crossover has poor data. Here's the source data confirming it:
It seems the gateway won't respond to my queries during the cutover, for about 5 seconds. That must mean that the gateway has a processor that prioritizes what it has to do. (Well, as long as its web server doesn't get hacked to do something electrically bad, we'll be OK! I'm realizing it may use the same code base and code processor for both web queries and logic; that's scary. As long as it's hack proof, then OK (!).)
By the way, the 18 seconds late timing is the timing of the Gateway, not of the samples from the meters, proving the Gateway is 18 seconds after UTC and therefore not GPS timing (I just confirmed my system was OK with ntpdate):
This zoom in to the 11:30PM transition shows the timing better:
This zoom in to the last hour also shows the later timing better:
I'm having trouble believing the 11:30PM crossover skips so many datapoints. Zoomed more in plotted as points:
As you can see, my sample rate is every half a second, yet the crossover has poor data. Here's the source data confirming it:
Code:
mysql> select meter,last_communication_time ,instant_power from meter where last_communication_time >='2017-08-19 6:30:17' and last_communication_time <='2017-08-19 6:30:25' and (meter='grid' or meter='battery') order by last_communication_time ;
+---------+----------------------------+---------------------+
| meter | last_communication_time | instant_power |
+---------+----------------------------+---------------------+
| grid | 2017-08-19 06:30:17.089413 | 5.760009765625 |
| battery | 2017-08-19 06:30:17.090362 | -1580 |
| grid | 2017-08-19 06:30:17.587284 | 13.670013427734375 |
| battery | 2017-08-19 06:30:17.590148 | -1610 |
| grid | 2017-08-19 06:30:18.085168 | 14.19000244140625 |
| battery | 2017-08-19 06:30:18.090217 | -1610 |
| grid | 2017-08-19 06:30:23.315487 | -1692.3899841308594 |
| grid | 2017-08-19 06:30:23.808858 | -1693.6199951171875 |
| grid | 2017-08-19 06:30:24.360047 | -1693.97998046875 |
| grid | 2017-08-19 06:30:24.858157 | -1718.4800720214844 |
+---------+----------------------------+---------------------+
10 rows in set (0.00 sec)
mysql>By the way, the 18 seconds late timing is the timing of the Gateway, not of the samples from the meters, proving the Gateway is 18 seconds after UTC and therefore not GPS timing (I just confirmed my system was OK with ntpdate):
Code:
$ date;curl -s http://redactedipnum/api/meters/aggregates|json_pp|grep last_communication_time
Sat Aug 19 10:25:25 PDT 2017
"last_communication_time" : "2017-08-19T17:25:25.894380607Z",
"last_communication_time" : "2017-08-19T17:25:25.891511928Z",
"last_communication_time" : "2017-08-19T17:25:25.89422894Z",
"last_communication_time" : "0001-01-01T00:00:00Z",
"last_communication_time" : "0001-01-01T00:00:00Z",
"last_communication_time" : "2017-08-19T17:25:25.891511928Z",
Last edited:
Ulmo
Active Member
I thought of some possibilities:
- During the crossover, the meters are taken offline to avoid trouble, such as logic errors from spikes, or other anomalies.
- The meters are online but can't believe what they're seeing, so they wait to be sure. I don't believe this, because I've seen spikes recorded by them before (such as when I turn on or off some huge un-ramped load).
- They're trying to only report non-spiked numbers. I don't believe this because I've recorded spikes from them.
Ulmo
Active Member
Ok, new information, or I forgot:
To me it is becoming crystal clear that PG&E is incentivizing us to use energy at the dirtiest times of day and night, and disincenting us from using it when they have the most power available and have the most clean power available, during the day. PG&E has a lot more power available mid-day, yet they don't let us use that energy because it is at super-peak rates then. Instead, when PG&E is starved for energy at 7PM on weekends, that's when we are incented to start using their energy, big time. It's backwards. Completely backwards. Now, with the PowerWall and family over, I'm seeing it first hand. I'm trying to balance my cost needs with my environmental needs: for instance, I charge up and try to do a small amount of discharge in the evening, but it's really hard to manage with incomplete software from Tesla, backwards time of use plans from PG&E, and a total lack of real time market pricing between the utility and my energy appliances (PowerWalls, SolarEdge inverter, solar panels, hot water heater, refrigerator, etc. etc.).
- I was at 70% state of charge. The app was set to self-consume with 70% reserve (reserve is for backup purposes). Right now, I have solar coming in, and home use exceeds solar. The gateway is not charging the battery, and not using the battery, and home is getting its power from grid. But:
- Yesterday, the gateway was charging the battery from solar when home use exceeded solar; home was coming 100% from grid, and battery was being charged 100% from solar. The difference? I had the "reserve %" higher than the state of charge. So, at 10:28AM, I just made the "backup reserve %" in "self-powered mode" set above the state of charge of the battery; if I'm right, this will cause the battery to charge from solar.
- I'm writing this post at 10:28AM. I'm going to wait 3 minutes to see if it starts charging the PowerWall from solar while home uses grid. (I'm on PG&E EV-A NEM2 rate plan, which means I'm off peak right now Sunday until 3PM.)
- At 10:30:38AM, the crossover happened: now charging battery from solar, and home using grid.
To me it is becoming crystal clear that PG&E is incentivizing us to use energy at the dirtiest times of day and night, and disincenting us from using it when they have the most power available and have the most clean power available, during the day. PG&E has a lot more power available mid-day, yet they don't let us use that energy because it is at super-peak rates then. Instead, when PG&E is starved for energy at 7PM on weekends, that's when we are incented to start using their energy, big time. It's backwards. Completely backwards. Now, with the PowerWall and family over, I'm seeing it first hand. I'm trying to balance my cost needs with my environmental needs: for instance, I charge up and try to do a small amount of discharge in the evening, but it's really hard to manage with incomplete software from Tesla, backwards time of use plans from PG&E, and a total lack of real time market pricing between the utility and my energy appliances (PowerWalls, SolarEdge inverter, solar panels, hot water heater, refrigerator, etc. etc.).
Attachments
Last edited:
bkp_duke
Well-Known Member
I can easily surpass 10kW during peak of the day sun. By the time I got to peak of day, I should have already fully charged the Powerwalls. If Tesla does something stupid like shut down my exporting of excess to the grid, I would be super pissed.
Jason's system is fundamentally different than the AC PowerWall + Solar. His system uses solar charge controllers direct to the battery. When his batteries get full, the charge controllers just curtail the solar by not letting the current through to the battery. The panels just float up to the open circuit voltage. In his system the inverters only supply AC loads from the battery. There is no such thing as "turning it down". His inverter system (Outback Radian) also has a grid connection that is pure battery charging and cannot feed back into the grid. He is not using the grid interactive connection on the Radian. The grid connection is only used when the batteries get low after prolonged low solar generation paired with high consumption. He curtails orders of magnitude more energy than he takes from the grid.
bkp_duke
Well-Known Member
Grid power got tripped at my house over the weekend. I guess it must have been something weird on the line as the entire thing shut down for about 10 seconds and then I got flipped onto battery power. Usually, when I test by flipping the disconnect at the meter, the backup gateway switches over very quickly.
Anyway, in that case, my batteries were near 100% and I saw a fault indication from my solar inverter complaining that the AC frequency was too high. Thinking about it, that seems intentional on Tesla's part, to shut down solar production when the batteries are full since there's otherwise nowhere to send the roughly 3.5 kW coming in and not getting used by the house. But it was kind of weird.
I suppose I'll be asking Tesla about that one, too. That said, it was nice having power for the half-hour or so that nobody else did.
Anyway, in that case, my batteries were near 100% and I saw a fault indication from my solar inverter complaining that the AC frequency was too high. Thinking about it, that seems intentional on Tesla's part, to shut down solar production when the batteries are full since there's otherwise nowhere to send the roughly 3.5 kW coming in and not getting used by the house. But it was kind of weird.
I suppose I'll be asking Tesla about that one, too. That said, it was nice having power for the half-hour or so that nobody else did.
Exactly! Which is why I was telling them they were nuts trying to make go to three PowerWalls.
