I did the home use < solar supply + battery supply && home use > battery supply && grid use = 0 test. My initial conclusion is inconclusive.
Step 1: I waited for a sunny day with decent solar output.
Step 2: I started washing clothes to set up for the dryer.
Step 3: I noticed someone else started the dryer for me. This put us at about 6kW use.
Step 4: I turned on two heaters. This put us at about what I thought was 9kW use (see below for actual).
Step 5: I turned off one PowerWall. This would reduce our battery capacity to 5kW continuous.
Step 6: I turned off PG&E breaker.
I then went to observe.
A: The power to the dryer was off. This is the first failure we've had in the backup gateway. I'm pretty disappointed. Power was on when I tried lamps.
B: I turned the dryer back on. The power dropped again. Lamp off.
C: I realized I was overtaxing it somehow, so went to turn back on my turned off battery.
D: Power was on. I also turned back on the dryer.
Now, time for postmortem:
It is possible I didn't do the math right. Solar was around 2kW to 3kW. 2kW+5kW=7kW, and that's less than 9kW. Let's see what's actually going on:
Let's go through the steps corresponding to the above graph:
Step 1: I waited for a sunny day with decent solar. Brings us to Sunday. Actually: 3,100W solar input.
Step 2: I started washing clothes to set up for the dryer.
Step 3: I noticed someone else started the dryer for me. This put us at about 6kW use. Actually: Black dots at 10:46-10:48 at 6kW use.
Step 4: I turned on two heaters. I thought this put us at about 9kW use. Actually: Black dots between 10:49 and 10:50 at 7,200W use.
Step 5: I turned off one PowerWall. This would reduce our battery supply capacity to 5kW continuous. Add this to the solar, and that's 8kW available.
Step 6: I turned off PG&E breaker. The noisy red line on the left is PG&E. Since the red line stops, that means that's when all recording stopped, and that's when power was lost the first time, shutting down my logger. As soon as the noisy red line disappeared is when I turned off the PG&E breaker.
Right at 10:50, it looks like the dryer was off but everything else was functional. I bet the dryer was off due to its thermostat, but I don't know. Then, when the dryer came back on, it went to ~7,200 watts again, and that was more than the Tesla Backup Gateway would allow, so it shut off. Interestingly, our dryer will shut off and stay off when power is off, so when the Gateway auto-restored, it could do so without the dryer load on. While this interrupts all electronics inside the house, if they are able to reboot unattended, at least they will come back to power. This is a mode I should consider making more robust.
All my logging came back online at 11:01 (I haven't yet automated this, so I had to manually restore a lot of it). It clearly has the sign of 0 utility (red dots) on the right, so the noise in utility (red dots) on the left that go all the way up until no points mean that utility was connected until all power dropped.
There are some possibilities.
- There may have been a glitch caused by one battery being turned off. This is the first time I've ever done that. That might have caused a variety of things to happen that I cannot determine today and don't want to test thoroughly.
- Gateway might have been in limp mode due to one battery being off. It might have refused to function to its full abilities.
Other than that, I had 5,000 Watts of battery available + 3,100 Watts of solar available = 8,100 Watts available, and only 7,200 Watts of use, so needless to say, if that were the full story, I'd be thoroughly disappointed in the performance of the Backup Gateway in this running mode. But, that's not the full story. We have to look at the balance of the two legs of the load and how the inverters are able to split it. Can they split it in whatever way we use it?
The dryer was on 240, so fairly balanced load. The two electric heaters were all on one side of the neutral median, so they were a lopsided load. So: 6kW of dryer taking all 5kW of the turned-on battery + 1kW of the solar, then 2kW of solar left in 240 mode so I don't know if that's only 1kW available on one leg of 120 VAC or 2kW available, so my 1,200 heater load might have been 200 Watts (plus other sundry items in the house) more than that one leg could handle, or it might not have been. This might have actually been a failure due to excess use on one leg of power. I'd like to know how to measure the different legs in the outputs other than buying more meters; obviously, the Gateway doesn't expose that information in the logging that I do; it just shows an aggregate number. The meters do measure the legs separately.
One thing I could do is put a meter on one of the legs to know the draw, and attempt to balance and/or imbalance it.
Now, I need to set up a test where I can draw ~12,000 watts of power. I'd struggle to come up with that. It could be done, but it would make the house a sort of noisy strange place for an hour.
Test 2:
Restoring power and solar during a power outage:
That works fine. I never turned utility power back on once I started this test. As soon as I placed the second battery in the ON state, with both batteries ON (normal state), everything started working again. Right now, the solar is charging the batteries. At least all two batteries I do need; utility power I don't need.
I hope Tesla's PowerWalls come down in price; they'd be a lot more beneficial at a price point that allowed homes to install 4 to 6 batteries, as opposed to 2. Some homes would want 8 or more, due to car charging and other concerns. Off-grid is a bit trickier with cloudy days.
Test 3:
Letting solar fully charge the batteries and then see what happens when there's more solar and no grid.
We'll see.
For those conducting these tests, it helps to have more meters available (or time to manually measure each pertinent piece of information), more logging (or time to measure each pertinent piece of information manually), more backups on the loggers (or time to manually take down manual measurements between each step), but regardless of all that, more time between each discrete step of the test. That makes the logging easier to decipher and read. I continually forget that; I am very fast when I test things, and it's hard to see what was going on in the logs afterwards when everything happens in one flurry.
So, what I'm really looking for now is a way to open up all the boxes, get out my manual tester, and then slowly perform the test with a bunch of manual measurements, doing things slower and more carefully. That will have to wait for a rare day that I am home and everyone else is working, but that does happen; perhaps within two months.
I'll see if I can sneak in a test today in case everyone leaves briefly later on. I'll leave both batteries on for the next test. I never turned PG&E back on; I have no need for it. And, I'll have to find some 240VAC load that can add to the dryer to put me at around 12kW. I'll probably need to buy a dual breaker to put into the box and a socket so that I can do that. I wonder what item I could use; perhaps rent something from a rental place.