“Just promise me it runs when the snow melts,” the customer said. He pointed at a pallet of solar panels in the gravel drive and asked a question I hear more often than I expected: “Can you put bifacial solar panels on roof?”
The short answer was yes. The correct answer was, “Not the way we nearly wired them.”
I’m a quality/compliance manager for a New England solar-plus-storage company. I review every system design before it gets installed—roughly 200 small commercial and residential systems a year. In 2024 I rejected about 9% of first-round equipment deliveries because data sheets or certification labels didn’t match the physical units. This story is different. The Pylontech UF5000 battery and the Epever 60A MPPT solar charge controller did their jobs. My array design was the weak point.
The small order I almost dismissed
This project started with a customer who owns two off-grid hunting cabins. He wanted one battery, no grid tie, and enough solar to run lights, a fan, and a small water pump. Total parts cost was under maybe $4,000. On a sales spreadsheet, that’s not much to write home about. We still took it, and I treat it like a large system, because a small system can fail exactly the same way a big one can.
I’ll be blunt about my bias: I have been the small customer many times. The vendors who answered my $200 questions are the same ones I later sent $20,000 orders to. Today’s small customer is often tomorrow’s repeat customer—and even if that weren’t true, a tiny cabin fire is still a fire.
What was on the bill of materials
For a Pylontech battery USA install, the setup was pretty standard on paper. The customer asked for a Pylontech UF5000 battery—a 5 kWh-class LiFePO4 cabinet with a BMS, in the same family as the US5000 and other low-voltage Pylontech modules. He paired it with an Epever 60A MPPT solar charge controller to handle solar harvesting. I’ve seen that controller used in hundreds of off-grid builds, and it’s a solid piece of equipment when the numbers are done properly.
The battery was not the risk. The controller was not the risk. The risk was the array behind the controller, and the panels in that array were bifacial.
The customer had bought six bifacial modules from a surplus dealer. The original spec sheet we quoted was for a different standard panel, but the dealer said, “It’s the same power, same voltage, basically the same thing.” I assumed that too. I didn’t redo the cold-temperature voltage calculation. That was the mistake.
The morning the sun exposed my shortcut
Commissioning morning started clean. The Pylontech BMS woke up, the Epever controller recognized the battery, and the array was producing by mid-morning. Then I left the site to check a separate job. About an hour later, I got an alarm notification: PV over-voltage.
By the time I got back, the Epever controller had stopped charging and the Pylontech battery had opened its internal disconnect. That’s what good safety systems are supposed to do. The problem wasn’t the lithium chemistry. The problem was that the cold morning string voltage—made worse by bright snow reflection around the panels—was higher than the controller’s absolute maximum input voltage. I measured it with my multimeter, then pulled up the datasheet on my phone. I had broken my own rule: never trust “same specifications” without verifying the actual model.
LiFePO4 battery safety precautions I now apply
LiFePO4 is a stable chemistry, but stable chemistry doesn’t resist wiring mistakes. Here are the LiFePO4 battery safety precautions I keep on the wall above my desk now:
- Recalculate open-circuit voltage at the coldest site temperature. Solar panel voltage rises when panels are cold. Bifacial panels add reflected light and sometimes extra current. If you don’t do the temperature math, the MPPT controller or inverter can see voltage it was never designed to handle.
- Torque every terminal to the manufacturer’s specification. Loose DC terminals cause heat, and heat is what starts the failures that people incorrectly blame on the battery. A battery can be perfectly safe and still be unsafe because the lug next to it was loose.
- Don’t rely on the BMS as the only protection. The Pylontech BMS is good, and it did exactly what it was supposed to do in this project. But it is the last line of defense, not the first one. Use proper DC-rated fuses and disconnects, and verify polarity before you connect anything.
- If the solar panel model changes, stop and re-review. “Basically the same panel” does not mean the same cold-temperature voltage, the same bifacial gain, or the same safety margin. The paperwork has to match the hardware.
- When using bifacial panels, leave room for the rear side. On a roof with tight clearance, the back side can overheat and the reflection gain drops. Bifacial modules need air and light behind them, not a hot shingle six inches away.
So, can you put bifacial solar panels on a roof?
Physically, yes. People do it every day. The better question is whether it makes sense on your specific roof.
On our project, the roof was a south-facing asphalt shingle roof with poor rear clearance. That is close to the worst case for bifacial solar panels. The back side of the module barely saw meaningful reflected light, and the lack of airflow increased heat buildup. The customer wasn’t wrong to want bifacial hardware; he was wrong for that mounting location.
If the same customer had a white TPO commercial roof, a bright metal roof, or a ground mount with open space and light-colored ground cover, bifacial panels could be a completely different story. I can only speak to our context: for a small off-grid cabin with a dark shingle roof, the standard panels were the right call and the bifacial panels were an expensive risk.
We reworked the array, corrected the voltage calculations, and upgraded the fuse and disconnect placement. The Pylontech UF5000 battery reconfigured cleanly, the Epever 60A MPPT solar charge controller resumed charging, and the system has been running without alarms since we restarted it.
Quality doesn’t check the invoice amount
I keep thinking about that morning—not because the equipment failed, but because I nearly let a small order get a smaller version of my attention. The Pylontech battery and the Epever controller both did what good components are supposed to do. The weak link was the person between them who assumed the panels were “basically the same.”
Since that install, the customer has asked us for a quote on two more cabins. That is the nice business ending. But the real lesson is simpler: every system gets the same voltage calculation, the same torque wrench, and the same uncomfortable review. Small customer or not, the cure for a bad assumption is the same checklist.
