Installation of Pololu's 305 kW solar generation system

This is the second post in a series detailing our experience over the past two years installing and operating a 305 kW rooftop solar system on our building in Las Vegas, Nevada. In the first post, I covered some of the background starting in late 2022 and how we committed to the $650,000 project by the beginning of January 2023, with a target completion date of May 31, 2023. This post covers how the actual installation went.

You may have noticed in the rough layout in the proposal that the 630 solar panels were not laid out in a simple rectangle. Our roof had 53 air conditioners, eight swamp coolers and various exhausts and vents and other equipment on the roof before we even started.

Pololu roof before solar panels, with 53 air conditioners, 8 swamp coolers, and various vents and exhausts.

Once we committed to the project, we had to finalize the actual layout, including sufficient space around all the equipment to service it (and limit shadows on the panels). We also wanted to prevent contiguous solar panel sections from becoming so large that they would force really long detours when accessing various parts of the roof. This was the final layout in the permits, with 3-foot buffers around equipment in yellow:

Pololu final pre-installation solar panel layout proposal.

Diagram showing ballast block requirements for one section of solar panels.

The solar installation company spent the first few months of 2023 making the specific plans and getting permits, including things like calculating the number of ballast blocks needed on our solar panels to keep them from blowing away. The total number came to 1,464 concrete blocks, each 4×8×16 inches and weighing 26 pounds, which comes to a total over 38,000 pounds or 17 metric tons.

We did our bit of preparation by having extra roof inspections and repairs ahead of the installation since any repairs would be much more difficult once the roof was covered (and we could also better know if the solar installers had caused any roof damage).

The original target date for start of installation was mid-March, but that got moved to April 17. With an estimated installation time of 3-4 weeks, completion by the end of May still seemed possible.

Pallets of solar panels started arriving in April and took up a good chunk of our warehouse.

The first several days of work were kind of anticlimactic from a spectator perspective as the workers measured out and marked the actual panel locations on the roof. The question mark on the exhaust on the lower left of this picture probably represented one of many times where the reality on the roof didn’t match what was in the plans:

Note also how the various connections to the HVAC equipment can expand the actual keep-out areas quite a bit.

But the exciting first crane day to get the panels on the roof came soon enough.

630 solar panels staged for hoisting to the roof.

17 tons of concrete ballast blocks going on the roof.

It was kind of funny seeing that pallet truck dangling from the crane and then on the roof, but I was kind of nervous about it. And, sure enough, they did make a hole in the roof with it. It’s easy to stack a lot of blocks on it, and all that weight gets transferred to just the three points of contact. Meanwhile, the roof is made of relatively thin OSB.

Hole in the roof.

You can’t tell from the picture, but that is really a fist-sized hole straight through the roof, and other than some fluffy insulation there is nothing between that and the warehouse floor 30 ft (9 m) below. (November 26, 2024 update: Robco requested that I point out that they did cover fixing it right away.)

With nearly two dozen workers on site, the solar panels arrays got assembled pretty quickly. The following pictures are from April 27:

On later days, they switched from the crane to a telehandler, which I think they found easier to unload and cheaper to rent or with better availability. These pictures were from May 1 and 2:

Using a telehandler instead of a crane to get the solar panels on the roof.

This was the progress on the roof by May 8:

Meanwhile, electricians were also putting up conduits inside and installing the three SolarEdge SE120KUS inverters.

Things seemed to be going relatively smoothly, maybe a little delayed, until May 20, which was the day to connect the system to our existing building power system. That was the first Saturday after construction started that we could get scheduled for NV Energy to come disconnect power from our building.

Disconnecting the 12,470V power connection from the street to the transformer for our building.

750 kVA, 480V transformer with 12.47 kV inputs disconnected.

With the power disconnected, the electricians needed to install a new 600A disconnect that would allow them to connect and disconnect the solar system from the power grid. The new disconnect needed to get connected to the line side of our main 2000A breaker, on the left in the picture below:

Main 2000A circuit breaker.

The new 600A disconnect is on the left in the picture below, and you can see the large conduit from that running along the ceiling:

Main electrical equipment, with new 600A disconnect for solar system on the left.

The line side is what is connected directly to the output of that transformer outside, meaning it is live any time there is power coming from the street and through that transformer. In other words, power comes in on the line side, through the 2000A circuit breaker, and from there to everything else in the building. The solar contribution has to be from the line side since that 2000A breaker protects the rest of the distribution system (if, say, 1900A were coming through that breaker from the utility and our solar system added another few hundred amps, over 2000A combined would be going into the rest of the system and could possibly damage something). There are probably other reasons the solar system is supposed to connect on the line side.

Line side tap called out in electrical permit.

Well, the electricians made the connection to the load side. Which they found out after power was restored to the building and the new 600A disconnect did not get that power until that 2000A breaker was turned on.

Take two, six weeks later

That’s right: we couldn’t get NV Energy back out on a Saturday until July 1, six weeks after that first failed attempt and now well into summer and beyond the point the whole system should have been running. I had gotten to know the electricians more by now, and I got a closer look at where the wires were attached.

Wires definitely going to the line side this time.

Without context, it’s difficult to appreciate how big these fuses are.

600A fused disconnect with fuses installed.

Anything over about 24V is “high voltage” to me, and I’m also not used to working with hundreds of amps. Here is that fuse in my hand.

600A fuse.

I was so amused by how big all these fuses and disconnects and wires were that I did not stop to wonder if they were big enough. Did you wonder if those wires could really carry hundreds of amps? Did you notice in this drawing from before that it calls for “(6) 500 KCMIL AL” wires?

Line side tap called out in electrical permit.

Fortunately, the inspector noticed.

Unfortunately, not only was this second attempt at connecting to the power grid another failure, but the oversight extended beyond the connection between the 600A disconnect and the line side of the main breaker: these wires I have circled here also needed to be doubled up:

Wires that also needed to be doubled up.

It’s really easy to draw three lines, or to write a label that says those three lines are six wires. But here is just part of those three lines in reality:

Wiring between 600A main disconnect and 200A individual disconnects.

Preparing the right number of wires to pull through the conduit.

The wires from the 600A disconnect I showed before come through that vertical conduit and then get distributed to three 200A disconnects that correspond to the three inverters (the third disconnect is off the picture to the left of the two shown). So most of that had to be redone. Those conduits are a lot more full now, and the gutters (the boxes at the bottom) are completely packed now.

Take three, July 22

Fortunately, NV Energy was able to give us attention on another Saturday just three weeks later. It was the same lineman from NV Energy and the same inspector from the county as the first two times so we were all getting to know each other by now.

With the wires doubled up, there was a new kind of risk we didn’t have before: that the wires would not be paired up correctly. These were huge wires coming straight off the transformer with no fuses or circuit breakers protecting them, so if some pairs of wires got mixed up, there would be a dead short with nothing to prevent a catastrophic failure. On top of that, the three phases coming from the transformer were still effectively shorted to each other through those transformer coils, so simple continuity checks on a regular multimeter were of no use since all the wires would read as connected to all the others.

Eventually we all convinced ourselves that the wires were definitely connected correctly, and we ducked behind our cars while the lineman reconnected power at the transformer. I’m only slightly exaggerating, and everything was boring and non-explode-y, just like we wanted it.

Line side tap with double the wires.

For completeness, here is that 600A disconnect with the wires doubled up. It’s made for it, so it looks completely reasonable in there.

600A fused disconnect with doubled-up wires.

And here are the individual 200A disconnects:

Final wiring of the three 200A disconnects for individual inverters.

And finally, at the end of July, we were done with all the physical connections.

Inverter installation as of July 31, 2023.

We passed our final inspection from the county on August 9. The electricians were wondering whether the inspector would even go on the roof, but he did, and he did some amount of looking around. (This was a different inspector than the one who had been there for the electrical room connections.) He did not check all 630 panels, but it seemed like he did at least a decent sanity check that nothing looked ridiculous, that what was there matched the basic plans laid out in the permits, and that spot checks of details did not reveal anything concerning.

That permit turned out to be just the starting point for starting some more paperwork with NV Energy. We and the solar installation company got the signed agreement to NV Energy on August 14, and it would take until September 20 for them to approve the project. During that time, they wanted various documentation of the work done and ownership details on the property. The information they wanted was probably reasonable, but it should not have taken five weeks to ask for it and verify it. We were not dealing directly with NV Energy, so it’s a little difficult to know for sure if it was NV Energy or the solar installation company that was holding things up, but the solar company did show us some emails to back up their claims that they were pushing as much as they could (I am inclined to believe them since we still owed them about 1/3 of the project cost).

The time was not all wasted, though, as the electricians still needed time to set up and troubleshoot the system. Here is the original plan again, followed by their marked-up picture of the final system:

Pololu final pre-installation solar panel layout proposal.

Notes from rewiring and rebalancing optimizer strings.

I am not sure if it was because of something like the strings not being balanced well or too much power going to one inverter, but on August 31, the inverter with the highest power (the green ones in the diagram above) failed. The electricians had been doing some intermittent testing but the system seemed to have been working, and the system failure was around 1PM on a sunny day with scattered clouds.

This is the NV Energy detailed power use chart for that day, on the old meter before they changed it to the net metering meter:

The light blue average usage line went to zero as the sun came up, and then started jumping around with up to 100 kW swings, presumably as clouds blocked the sun. This is how the inverter looked when the electricians opened it up:

The 200A fuses in the disconnect connected to that inverter were also blown.

Home stretch

Here are the remaining milestones as we finally got the system running:

• September 29: The left inverter sub-module and the bottom main controller unit were replaced on the failed inverter.
• October 3: NV Energy changed out our meter to the net metering version (one that could measure power flow in both directions).
• October 4: NV Energy sent us confirmation email with approval to operate.
• October 5: We got our first look at the SolarEdge monitoring site:

Next time, I’ll be covering our experience over the first year of operation. Did we hit our projected 528,000 kWh of production?