With the last of the panels mounted and the wiring done, the next step is, to a lot of people, the most-feared part of the process: inspections. You have to meet code.
And… at least out here, it’s not nearly as bad as I feared. It’s far more of a TLAR inspection (That Looks About Right) - though how much of this was that it was obvious that I’d taken my time and done things correctly, I’m really not sure. I’d rather not find out, either.
The National Electric Code
If you’ve not been buried in this sort of stuff for a while, it would be easy to think that the National Electric Code was put together by some official group of electricians maintained by the US Government. It’s not. The NEC is part of the National Fire Code series - put out by the National Fire Protection Association. This is a private trade group, not a government agency, and their focus is on keeping electrical systems from burning down buildings. Does this mean you end up with some comically absurd wire gauges at times? Yup. Tough. Deal with it, because it’s cheaper than paying a professional engineer to argue. Unless, of course, you’re someone like a power company, with your own insurance, at which point you can do literally whatever you want as long as your staff professional engineers are comfortable with it. If the run from your transformer to your meter is a couple sizes under what NEC would require? No problem, they’re not bound by NEC if they don’t care to be…
Most areas adapt some version or another of the NEC, but it’s up to your local jurisdiction what version they want to see. It changes over time, and for solar, it’s getting far, far more picky over time about things like rapid shutdown. I’m not going to say for certain that some particular microinverter monopoly might have bought themselves a handout in the NEC 2017 requirements (which now require per-panel rapid shutdown electronics), but… it sure looks that way sometimes.
In any case, if you’re doing your own install, just go buy a paper copy. You’ll need it, you’ll probably want to take some notes in it, and I’m sure it doesn’t hurt if you have a large, obvious, physical copy around while your inspector is poking at stuff. It’s also a pretty solid mediator if you have disagreements - flipping to the relevant page during an inspection for discussion indicates you’ve probably done your homework, if there are any questions - though it’s not your interpretation that matters, it’s theirs.
Seriously. Buy it. Read it. You’ll need it.
The Code Inspections
For most electrical installations, there are going to be a range of inspections. For my solar install, in my jurisdiction, I really only needed two inspections. I might have been able to get away with one, but the first one serves a useful purpose as an opportunity to ask the inspector about anything you’re unclear on or they might have a local understanding of - and to allow me to cover my trenches.
The first inspection I got was the “rough-in inspection.” For this, I had the bulk of the wiring done, conduit run, trenches dug, solar frame side combiner/disconnect boxes mounted, etc. The inspector verified that the equipment I had installed was the equipment listed in the plans (apparently this is a major issue - submit plans for one set of equipment and install something totally different), and generally went through the approved copy of the plans from the meter to the panel mounts. As far as I could tell, he was looking to verify that what I’d submitted was what I’d actually wired, and, in the process, looking over general workmanship.
Look. If you’re doing something like this yourself, you have all the time in the world. Spend the time to make it look good. Arc your wires around corners gracefully. Spend the extra couple bucks on the rounded end caps for your conduit terminations. Align everything. Make it look like someone who cared put it together, because, you do! It took me far longer than it would have taken an electrician who does this constantly to put the system together - but I also made it look better. The inspector, who sees all sorts of professional work, asked me if I’d done all of this myself (which I had), and observed at several points that it was an awful lot cleaner than the “professional” installs he has to deal with. This is a good thing. Impress the inspector.
One of the main values of the rough-in inspection (and a good reason to request one - you’ve paid for it in your permit fees!) is that you’ve got an inspector, onsite, available for questions. I don’t know how it works elsewhere, but out here, there’s an inspector that covers my area, and he’s the one who does everything. There were a few points I was slightly unclear on with regards to labeling, and I got clarification. I’d called him to ask about the finger guards for “low” panel installs (you either need to “guard the wiring” or fence the area in), and he wasn’t able to give me a great answer without seeing things - but, onsite, looking at it, we were able to discuss it and agree that some plastic fencing, stapled in place, would meet the requirement.
Be nice - they’re the one who has to sign off on stuff. But know what the requirements are, know what you’ve done, and be willing to defend your work. If you’ve met NEC requirements, and they ask about something, don’t be afraid to defend your work. During the rough-in, there was a question about an unsupported vertical run of conduit. The NEC requires conduit supports every 3 feet, and I was quite certain that it was far less than three feet unsupported (I’d placed the support based on this requirement). A quick check with a tape measure, and, yup. It was fine as it was.
Once you’re done with the rest of the install, schedule them back for a final inspection. For my install, this was surprisingly quick. The only major change for the final inspection was that the panels were installed and the conduit trenches were covered. I’d covered everything else during the rough-in. It went quick, with no corrections, and I didn’t expect to see him again any time soon…
The Power Company Inspection: Round 1
After the Department of Building Safety people made their inspection, it was time for the power company - who, as I’ve noted, doesn’t care about code. They care about their own things, such as having UL listed inverters, validating rapid shutdown behavior (and the time for them to come back online after connection), and generally making sure that you’re not going to hurt their system. Burn your house down, they don’t care. Damage their power grid somehow, they care. Don’t meet their (perhaps arcane…) requirements, well, that’s your problem.
And we immediately ran into a problem, because according to the guy from the power company, my disconnect didn’t meet their standards. It didn’t meet the required “conductors visibly moved out of position” part of the disconnect requirements, and therefore couldn’t be approved.
So, reasonably enough for the more than mild panic I was in, I asked where it said this. I’ve read everything they put out - I thought. So, we both set about trying to find the relevant part of the documents, and both came to the same document - mine in PDF form, his in printed form. And it said absolutely nothing about “visibly disconnected conductors.”
Weird, but… we both had slightly different versions of the same document, and it said nothing of the sort. So, on we went, the system passed their requirements (as expected), he put the required labels on, swapped the net meter base in, and I thought we were done with things.
Except… Schedule 68…
Well, we weren’t. A few days later, he came back. It’s never good news to see the inspector again after you thought they were gone, and he came bearing quite bad news: They’d found the requirement for visible disconnects (in their rate schedules - it never made it into any of the other documents, including the Interconnection Standards document), and, as I didn’t meet that, my system didn’t meet their standards and therefore couldn’t be approved.
I was using circuit breakers for disconnects. It met NEC, but it didn’t meet the (independent) requirements from the power company. And nobody along the process, plans review, inspection, etc, knew a thing about this.
Now, I figured there was a chance of some rework somewhere in the system, but I hadn’t planned on “Your entire side of house part doesn’t meet the requirements.” Because I thought I’d done my research. The problem here was that the plans review process took months, and I didn’t have months left to get things online. I was already in late October, and I had a hard December 20th deadline for 1:1 kWh for kWh net metering.
My system is designed for a post-net-metering world - but, even if I don’t agree with the implementation, the net metering subsidy is worth an awful lot of money. Because I’d optimistically bought panels back in 2019 (and therefore had made a “financial investment in the system,” even though it wasn’t online yet), I fell into window that would allow me to be grandfathered in to net metering for 25 years, if the system was online by Dec 20. But if I had to go through the full process from the start, that wasn’t going to happen.
However, the changes I needed to make were only on the AC side of the system. It wasn’t any change to the core of the power generation (panels, DC wiring, or the inverters) - just replacing some disconnects on the side of the house. A polite email to the relevant people later, explaining what was going on, and they agreed that my changes were just some alterations to an existing, approved system, on the AC side, which didn’t require new plans review or permits. Do the work, get another inspection, and I’m good. Phew. Definitely better than feared.
So, here’s the problem. This is my side-of-house system, with the very nice looking MidNite Solar combiners and disconnects. The right hand box feeds one of the 6kW inverters into my panel, and the left hand box combines the 3kW and other 6kW inverter into a single feed to the “bonus breaker spot” up top (a 50A, always-live slot at the top). The breakers in the right hand box serve no real purpose beyond being a disconnect (there’s a breaker in the panel that can cut power to that circuit), but the breakers in the left box are important. They protect the wiring between the house panel and the inverter. There’s the 50A breaker in the house panel (which is fine for the 8AWG between the house panel and the combiner), then some smaller breakers (15A & 35A) protecting a 12AWG run and a 10AWG run. I can’t just merge those, because a short at the far end would potentially put 50A onto a 12AWG run - which isn’t allowed by NEC. I’m back into NEC here, even though it’s power company requirements I have to meet. Always fun.
But, I eventually found some standard enough disconnects that would work for what I needed, and that I could find, if not locally, with a reasonable delay (about a week shipping - chewing into my buffer here, but… it’s still fine…). What’s important about these disconnects? Look at the center, red, rotating segment. See how the metal tabs on them are visibly not touching the tabs on either side, so the circuit is obviously not connected? This is what I need to meet the power company requirements. Not some crappy circuit breaker that might not work… I’m fuzzy on the rules, but it’s their grid, their rules, even though they seem a bit absurd at times.
For the right disconnect, which is just a straight through feed of 10AWG, I had to run another run of wire from the house panel, but I was able to reuse the run from the inverter (which was the hard one to pull). Why is the inverter run connected to the bottom? Because I left enough slack that I could connect it there. The ground was tight, though…
Now, you’re not supposed to cut conduit with wire in it for obvious reasons. And I certainly shouldn’t recommend doing something like that, if you happen to need a chunk of conduit run shortened and don’t feel like pulling the wires again. But I will suggest, purely from experimental experiments of an experimental nature, that a typical metal pipe cutter of the “twist around the outside and crank the cutter wheel down until it gets through” sort of cutter will cut Schedule 80 PVC without doing the slightest bit of damage to the wires inside. The wheel never actually penetrates to the inside, but makes the remaining wall thin enough that you can snap the conduit chunk off and be fine. Measure carefully.
What I did, in the end, was to move my combiner with the circuit breakers up, leaving room for a disconnect squeezed in below. Don’t ask hard questions about the weird placement of the assorted conduit joiners. This was done during a time when I couldn’t find much of anything I needed, and the new “L” conduit boxes for the combiner were bought on eBay, because I just couldn’t find them locally. The End of the Year of the Pandemic (#1) was really, really rough for things like PVC conduit bits and pieces.
But, another day or so of work, and all of it was done.
Yet More Inspections and Online Again!
And, therefore, I scheduled Yet More Inspections. Another NEC inspection (the inspector thought this somewhere between absurd, entertaining, and a waste of time, but was going to pass the details up to the plans review types, so they could mention in the future that a particular plan was NEC compliant, but not power company compliant). And another power company inspector (from an inspector who was somewhat irritated that this was his third time out here - not at me, but at the fact that a perfectly safe and operable system, fully labeled, was rejected for what he and I both agreed were pretty well nonsensical reasons). But, new labels applied, new paperwork signed, and, I’m good to operate! Again. Right as winter hit.
With that… things got quite boring. I started learning how the system operated, it sat in the sun (or clouds, or inversions) and produced power - sometimes a lot, sometimes nearly nothing.
But they’ve been working throughout the winter. I ran the meter forward about 3.1MWh over the winter (meaning I pulled 3.1MWh more from the grid than I pushed), because we heat with electric - everything in the house is electric. Once spring hit, it started pushing rather substantially more to the grid than we use. I’ve exported 70kWh some days so far (on production of 95kWh or so). And the meter has been run back well past zero (it rolls over at 100,000 kWh, so this is 400 kWh net exported, after rolling back winter).
Finally, just how much of a difference does the whole “east/west facing panel” thing make? I’ve got some clean production curves to share to show the difference! These are from early June, which is roughly the same as summer solstice (about half an hour of sunlight difference, but this happened to be a clean day).
We’ll start with the south facing panels, because this is what everyone is used to seeing for a solar curve. On a clear day, you find a nice smooth curve, starting as the sun comes past east, and extending until the sun goes past west (beyond those points, it’s just lighting the back of the array - the long low tails are from ambient illumination). Simple, normal, and this is what most systems produce. But my system isn’t normal. I went out of my way to put in some panels that face east and west and prattled on about a “Long Solar Day” or something. What does that mean?
To demonstrate the difference, this is what the curve looks like on my east array (which is higher and, perhaps obviously, to the east). As soon as the sun comes over the slight hill to the east, production whomps right on up. No graceful start here. By 8AM, as the south panels are getting started, this frame is passing past 2kW, and keeps on climbing up. There’s still a mid-day peak when the sun is overhead and both sides are illuminated, but notice how the curve is much flatter and longer. On the right, in the evening, there’s a steep drop that happens when the array starts getting shade from the west set of panels, and only generates off diffuse illumination.
Moving west, the west array is a bit lower than the east array, so it’s shaded longer in the morning. I wasn’t too worried about this, as demand is far higher in the evening (in the house and on the grid, especially in the peak of summer when air conditioning is running). There’s a nice smooth drop in the evening, but it’s holding on and generating power until 9PM!
And, finally, proving that I’m not going to find a second career as a graphics designer right now, I’ve overlayed the curves on each other. “Full deflection” on the charts is related to the inverter capacity, and since my south facing panels are on a 3kW inverter (vs 6kW for the other frames), I’ve scaled the production curves to match each other here. I won’t say the south panels do nothing, but… they round out system production numbers and were intended to help with EV charging during the middle of the day. As it turns out, I probably could have left them off and been perfectly fine.
You can see the “cut” on the left is larger than the same cut on the right, which reflects the height difference in the arrays.
But, above, it’s not really a fair comparison - the east/west frames have 4x the panels of the south facing ones (24 vs 6). So, to compare more directly, I’ve scaled up the south facing panel production by 4x (to simulate 24 panels) and overlayed it with the other arrays. Now the extended solar day really shows up! By 8AM, both A-frames are up past 2kW - where if they were south facing, they wouldn’t hit 2kW until about 9:15. Mid-day production is higher with the south facing panels, as expected, but in the evening, by 6PM, they’re dropping hard. In the 5-8PM range, when the south facing panels are dropping rapidly, the As are still producing strong, and we continue exporting into the evening. This is designed to help prevent duck curve issues, and with our consumption patterns, it certainly helps.
To pick a particularly clean recent day, this is what our hourly exports to the grid look like on a day with no exceptional use. Notice there’s no evening duck, even though we use power in the evening - it’s just a graceful ramp down of exports to our somewhat stable overnight usage.
Per-Panel Production Numbers
One of the criticisms I’ve gotten about this system design is that I’m “wasting panels” with the As - they don’t produce as much as the south facing panels. And it’s entirely correct. In the winter, with the sun rising in the south, the As don’t produce as much per panel as the south facing ones (though they still produce at more useful times). But how about the peak of summer, when they’re able to grab the northeast morning sun and cling to it until the sun sets in the northwest? Using the numbers from these set of curves:
The south panels produced 10.44 kWh on 6 panels, for 1.74 kWh/panel.
The east frame panels produced 42.58 kWh on 24 panels, for 1.77 kWh/panel.
The west frame panels produced 41.89 kWh on 24 panels, for 1.75 kWh/panel.
So, during the long solar days of summer, the panels on the A-frame actually produce just as much (if not slightly more than) the south facing panels!
Winter is certainly far worse on that metric. Our maximum production in January led to north/east/west panel production of 1.76/0.79/0.82 - quite the difference from summer when they’re dead even, with the A frame panels producing less than half of the south panels. For an off grid system, the long solar day is valuable, and even for a grid tied system, the timing of production is valuable, but they do produce less per panel, annually, than if they were all facing south.
How does it average out? With a bit over half a year of data, total kWh production per panel (north/east/west) is 267/191/187. So a very real difference, but we’ve still not hit peak summer sun yet. I expect this will pull closer to even over the rest of summer and into the fall.
Thus Ends the Solar Saga!
So, everything’s up, online, and working. It’s been an interesting ride!
Actually, I have a few more posts about solar, and then a summary in the same style as I did for my office. But I may post those later if I have more interesting stuff to talk about in the meantime.