An Enphase Ensemble Installation

[svetz]

... meter supports zigbee...

Previously, there were two strong Zigbee channels at 21 and 25, for the IQ Battery (evident as those signals disappear with the two dongles unplugged), AFAIK that's the only zigbee device I can pick up. It was also determined a while back and there have been multiple new firmware releases and reboots... so it might have changed. As you can see to the right, signal strength to the batteries is quite good today (possibly the last patch?)

Back in #390 I tried changing my router's channel to avoid interference as the WiFI & ZigBee bands overlap. My neighbor's WiFi are on channels 1, 6, and 9 (you can see these from a smartphone app like WiFI Analyzer). Changing it didn't improve the signal or fix the problem (occasionally the battery loses comms, which is silly given how close together they are, the ZigBee dongles now live in a box between the IQ batteries and Enpower, see #401). From the data below, the right side of the spectrum is being used for ZigBee, so my router near the IQ batteries is set to channel 3 and the far one is set to channel 10. Channel 10 isn't ideal, but it's not changeable in that router and only slightly overlaps 0x14. It's -84 dB near the ZigBee from that router, so hopefully lost as noise.

Looking for ZigBee Traffic

Channel	Traffic	Channel	Traffic
0x0B		0x13
0x0C		0x14 (20)	varied data
0x0D		0x15
0x0E		0x16
0x0F		0x17
0x10		0x18
0x11		0x19 (25)	Beacon mostly, 45b
0x12		0x1A	802.15.4 ACks (keep-alive?), zigbee data on event?

Looks like Zigbee 0x15 moved to 0x14. I didn't see any traffic on the other channels, but I only spent a minute per channel. I'll try again at 10 min per channel. Wonder if there's a Zigbee Analyzer I can just leave up and running.

Update: No joy with the 10 minute scans. Possibly it only broadcasts when it receives a signal?
Update 2: Learned the utility data is updated every 15 min via cellular, repeating Zignbee scans with 15 min per channel. Found traffic on 1A

 

[svetz]

Getting a new wholehouse surge protector, the old one died , don't forget to check your LED indicators!

 

[GXMnow]

Did it sacrifice itself to save you house from a real surge?

If you can't read the new power meter, can you get the data you need with a consumption meter like a watt-node at your main input? Then your not tied to their equipment.

 

[Hedges]

Here's the one I installed.

I don't know how much energy it is rated for, but one thing I like is MOV to clamp at 200V, and if that set blows a light comes on and another set is waiting at slightly higher voltage to take over.

Testing Metal-Oxide Varistors (MOV)

Surge arrestors are used on AC and DC side of PV inverters for protection, to clamp high voltage transients. One inverter I have (actually several) are in the SMA family SUNNY BOY 5000-US / 6000-US / 7000-US / 8000-US. These come with a kit of MOV to protect the DC side: The MOVs are about...

diysolarforum.com

 

[svetz]

Did it sacrifice itself to save you house from a real surge?

Probably not one super big surge, probably more like death by a million paper cuts. I thought to check as one of my monitors recently died (don't feel too bad, if was 14 years old and after seeing this sort of glad it did). Surges are very destructive to electronics and we seem to be blessed with them.

... can you get the data you need with a consumption meter like a watt-node at your main input?

I have that data already.

Then you're not tied to their equipment.

I like reading their equipment as it's the source of billing data. Being able to compare the two ensures one hasn't gone south. Sort of a lazy-guy's trust, but verify. From the operation's manual, there's also interesting data I don't get like quadergy. Suppose I could calculate it. It would be nice to minimize as it's just money thrown away and probably adds up over time. Should probably start a thread on it to see what others have done (although would like to avoid the apparent power issue, remember the frouhaha?).

Anyway, the power company is moving to put hourly data online and when they do that'll be the easiest way to validate it. They have the consumption data already, but not exported data.

Here's the one I installed.

Holy cow! How did I miss that thread???? Thanks for linking it in. Posted there, hope to see more discussion it.

 

[svetz]

3 of my 4 batteries drop from 60% to 33% in a short period as shown to the right. As there was no corresponding power through the CT, it's once again the issue of the Envoy re-syncing with the actual SoC. Disturbing as it seems, this is actually an improvement as before the prior s/w update they didn't resync until 10%. I've reset the default SoC back to 100% so they can top balance tomorrow. Then I'll then put them back to 60% to see if the last patch has finally taken care of the SoC between the Envoy and the battery getting out of sync.

​

 

[svetz]

Set the batteries back to 60% SoC this morning, it's been overcast/rainy the last few days.

Received an eMail from Enphase:

This is to inform you that you can continue to use Enphase API v2 until January 18, 2023, under the existing pricing and license agreement. Effective January 19, 2023, API v2 will be phased out and not available to any customers

Explored the old API but never used it as it was all remote. They have a new 4.0 API, I love they way they laid out the documentation. Very clean, easy, and detailed. I'm not opposed to remote monitoring for things that aren't critical when the grid/communications are out, but didn't see anything like that in that docs and even if there was, they already have tools (e.g., Enlighten) to explore that data. Those APIs mainly look like they're for installers to watch the health of their customers' systems.

 

[svetz]

Currently, my highest outputting panel is 11W due to rain and it's about 77F (e.g., STC). The ambient solar is currently 46.2 W/m²
What sort of efficiency hit do the panels + microinverter take at very low light?

From the solar meter:



The LG Neon panels are about 1.6m², so at 340W that's a (340W / ( 1.6m² x 1000W/m² =) 21.2% efficiency at STC.

46.2Wactual x 1.6m² = 74W possible. 11 Wactual/74 Wpossible = 14.8% efficient at 4% STC illumination. So, ~6% efficiency hit.

 

[svetz]

Called Enphase Tech support today as I realized that the Enlighten Live Status hadn't worked since the upgrade, it kept coming back as couldn't connect. They knew exactly what I was talking about so it must have happened to some percentage of folks after the last upgrade. They had it fixed right away.

 

[svetz]

One of my batteries jumped from 60% to 30% SoC over a 15-minute polling period with no corresponding discharge and it had a very strong wireless connection over that time. As it didn't look like the latest patch fixed that annoying bug (Battery SoC being reported higher than it is) I called tech support and the guy said he knew about the problem and it was a software bug recently fixed. Then he realized I was on the latest version and went off to talk to the engineers.

What happened next was odd. He said he'd made some changes and to keep an eye on it for 24 hours. When I looked, the low battery had jumped back to 60%. I had been assuming that the battery was slowly discharging and the Envoy wasn't getting updated. But it being 60% instantly seems to indicate that it wasn't really discharged. So, now I'm not sure what was going on. I'm going to let it sit as it is currently provisioned through tomorrow. But if it all looks stable after that I'm going to charge it up 100% and do a full discharge test.

 

[svetz]

One of the batteries blipped to 0% this morning, but the next cycle had it back to 60%. Normally that would be a communication fault, but it happened on one of the batteries with an exceptionally strong signal normally. Must have reset or something. I'll hold off another day before I run the discharge test.

 

[GXMnow]

Reporting an accurate state of charge is difficult on LFP batteries. The voltage changes so little that the only real way to do it is to track the amp hours going in and out, and it can only reset to full or empty when it charges or discharges into the knee at the ends. If the voltage stays in the wide plateau, the actual SoC will most certainly drift off from the reported SoC. The longer it stays inside the 20% to 80% range, the worse it can get. And this is true for any BMS or coulomb counter on LFP batteries. To keep it semi accurate, it needs to charge up to full once in a while.

Even on my NCM cells, which have a fairly linear voltage slope, the SoC reported by my BMS has varied quite a bit since I have only been charging to about 90%. I did a full charge once to nearly 4.2 volts per cell, and then when I was only cycling up to 4.05 volts per cell, it was reporting the correct 85% state of charge for a while, but then the BMS decided that since the charge kept stopping at that voltage, it started reporting that as 100% charged. I don't check my BMS reading very often any more, so I am not sure when it made the change. But since I added the DC charge controller to the system, it has pulled the voltage a little higher on a few very sunny days, so now it is back to reporting only 90% at 4.05 volts per cell again. With LFP cells, I know that SoC drift would be a lot worse.

I have to wonder if Enphase is having a computer program go over the charge and discharge data trying to get a closer estimate to load into the system as the current SoC?

 

[svetz]

In the Enphase IoT thread I said I needed to retest to see if there was any frquency shifting when the panels were throttled. Turns out there is! I probably didn't notice it before because the excess solar was probably recharging the battery, the battery needs to be near 100% before it decides to shift. So that's pretty cool, that means we do have what we need to calculate excess solar without other devices. Might not even need the frequency data, possibly if the battery is over 95% you could just say "yup, that's excess power available". I believe it's a linear % throttleing from 61 to 62 Hz, so 61.12 would be 12% throttling. I need to do this on a much sunnier day.

 

[Hedges]

SMA off-grid and backup default settings are 100% up to 61 Hz, linear ramp down to 0% at 62 Hz.

UL-1741-SA I think is 100% to 60.5 Hz, linear ramp to 0% at 61 Hz. As graphed, I think your inverters are being told to deliver zero power.

UL-1741-SA also has "must disconnect" after some time above the exact same frequency, 61 Hz. I'm not clear if that still applies for an inverter doing this linear curtailment. For AC coupled battery systems, you want some margin between "zero output" and "disconnect, stay offline 5 minutes." For SMA those are 62 Hz and 64 Hz, so large dead band. Frequency overeshoots a couple 1/10th Hz, so good to have a dead band before instant disconnect. (There is no slow disconnect in the SMA scheme, but UL-1741-SA has both delayed and instant disconnect thresholds.)

 

[svetz]

Thanks! Enphase use to have a graph in their AC Coupling document, but I only see the aussie version now.

Update: Seeing 60.1 to 63 Hz in Validating the test procedures described in UL 1741 SA and IEEE P1547.1. The Aussie version shows 50.1 to 52. If it starts at 60.1 Hz and goes to 63 Hz that would be 35% throttled in the picture above. Definitely need a sunnier day and probably an amp meter.

 

[svetz]

Updated the OP with November's production numbers.

 

[svetz]

Discharge Test
Discharged the batteries from 100% to 10% SoC and saw 12.1 KWh of energy, the batteries have 13.44 kWh new, so 12.1 ÷13.44 = 0.90003, so looks spot on. They're 2.5 years old so thought they'd be a bit less based on the datasheet's ">70% capacity, up to 10 years or 4000 cycles" which is about a 3% loss per year. Wonder how other's LFP installations are doing over time?

 

[Hedges]

Discharge Test
Discharged the batteries from 100% to 10% SoC and saw 12.1 KWh of energy ... so looks spot on.

How was 10% SoC determined?
By measuring 0.9 x 13.44 kWh = 12.096 kWh and calling that 10%? ?

 

[svetz]

How was 10% SoC determined?
By measuring 0.9 x 13.44 kWh = 12.096 kWh and calling that 10%? ?

SoC is usually measured via current flow through a shunt so that makes sense. Ideally, I'd get the pack voltage or run it to 0% to see if I get the full 13.44 kWh, but the lowest it can be set is 10% and the battery voltage isn't reported (guess I need to get back to work on cracking the zigbee ;-). On the upside, I did see 12.1 kWh, so I'm at least still ≥ 90% capacity and within the warranty. I wonder if anyone has taken an IQ8 card out to check bus voltage?

 

[svetz]

My little local power co-op just started showing my kWh exported on their online usage data. Matches up to the data from Enlighten very nicely.

Love that they have the temperature line as well, you can definitely see which days the AC didn't come on (e.g., 11/20)...