In August 2024, I switched off the mains at a small warehouse to test the backup system we had just installed. The battery was 87 percent full. Solar panels on the concrete roof were generating. The hybrid inverter showed no errors. Then the facility went dark.
I'd like to say that was the moment I learned my lesson. The truth is, I should have learned it two months earlier, when I designed the system on paper and treated a hybrid inverter like an ordinary grid-tie inverter with a battery attached.
I've been handling B2B solar-plus-storage orders for six years. I've made and documented five significant mistakes in that time, totaling roughly $28,000 in wasted budget. This one cost about $9,400 in rework plus a three-week schedule delay. I'm sharing it because the checklist we use now has already caught the same mistake on another project.
The project: a concrete roof, a stack of LFP, and one "simple" hybrid
A property manager called us in March 2024. Their depot had old PV panels on a flat concrete roof, and they wanted to add battery storage so the office, gate controls, lights, and one ventilation fan could keep running during power cuts. Nothing crazy—no need to run the truck dock or the EV chargers.
We specified the Pylontech Phantom S battery. The whole Pylontech range uses lithium iron phosphate cells—what some datasheets label an iron lithium battery (LiFePO4 if you want the full chemical name). I'm comfortable with LFP from years of using the US series, and the client liked that they could add another cabinet later without changing the inverter.
The rest of the bill of materials looked boring in a good way. A Pylontech combiner box sat between the battery stacks and the inverter, giving us a single DC disconnect and a clean way to terminate the parallel battery cables. On the roof, we used an aluminum rail product that the vendor listed as a solar mounting for concrete roof. It was straightforward: aluminum rails on stainless-steel anchors, no roof penetration through the membrane. The installation itself took five days, as far as I remember. Every function test passed while the grid was on. PV production went into the building, the battery charged from surplus solar, and the app showed the client's load curve during the day. Then came the one test that mattered.
Where the design went wrong
While the grid was on, every load in the depot was fed through the hybrid inverter's protected-load terminal. That is a natural thing to draw if you think of a hybrid as a grid-tie inverter: grid in, building out, battery and PV attached somewhere in the middle.
It is also wrong for backup, at least on the model we installed. In an outage, the inverter's backup output is limited by the power electronics, not by the battery's stored energy. The battery can supply plenty of energy for hours, but the inverter can only push a certain number of amps to the protected side. If the protected circuits draw more than that, the inverter shuts down to protect itself.
Our "critical loads" list—office, lights, gate, one fan—was about 2 kW. The full depot, however, also included an air compressor that started automatically every fifteen minutes or so. That compressor was on the same protected circuit. When I opened the mains breaker, the inverter took over for about ninety seconds, the compressor kicked in, and the inverter shut down with an overload fault.
I want to be clear that I am not blaming the equipment. The battery was healthy, the BMS communication was working, and the Pylontech combiner box did its job. The fault was my circuit design. I had not physically separated the non-essential loads from the backup side.
How does a hybrid inverter work? (the part I learned late)
I'm not an electrical engineer. I'm an installer and project manager, so I'll use the version I use with our own crews.
A hybrid inverter is basically a grid-tie inverter plus a battery inverter plus an automatic transfer switch in one metal box. The DC side has inputs for PV and for a battery. The AC side has two connection groups: one that goes back to the grid, and one that goes to loads you want to protect.
When the grid is present, the transfer switch keeps the protected loads connected, and the inverter can run in several modes: solar first, then battery, then grid. It can also export surplus.
When the grid drops, the transfer switch isolates the protected-load side from the grid on purpose. That is an anti-islanding requirement, not a design flaw. The load on that protected side is now supplied by the inverter's own output, fed by PV and battery. If the combined load is bigger than the inverter's backup rating, the unit simply cannot sustain it. The battery chemistry—iron lithium or anything else—does not change that limit.
The short version, if you ask me: the battery gives you time, the inverter gives you power. You need both. I have mixed feelings about this because hybrid inverters are genuinely useful. The problem was not the product class. The problem was that I sized the system around energy in kilowatt-hours and almost ignored the instantaneous power limit of the backup terminal in kilowatts. Those are two entirely different specs.
If you under-spec the backup side, the inverter becomes the most expensive relay in the building.
The fix
We went back to the site with a licensed electrician and split the depot's distribution into two sections. The compressor and the other non-essential loads stayed on the utility side. The office, lights, gate, and fan were moved to a small subpanel fed from the inverter's backup output.
The battery stayed exactly where it was. The Pylontech combiner box did not change. The Phantom S stack did not change. What changed was the wiring and, just as important, the explanation to the client.
When we repeated the mains test in September 2024, the protected loads never blinked. The compressor was off, so the inverter had no reason to trip. The client accepted the rule because we had documented which outlets would stay live and which would turn off during an outage.
What we check now
Since this project, every B2B quote we send includes one extra page: load classification. Here is the part of the checklist that would have saved this job:
- Separate every load into two lists before ordering anything: essential loads that must run on battery, and non-essential loads that can wait for the grid. If a machine can't be asked to stop, it belongs on the utility side by design, not on the backup side.
- Confirm the hybrid inverter's backup current limit from its manual, not from the sales brochure. Pay attention to starting current, not just steady-state watts.
- Physically open the mains breaker before handover. A simulation in the app is not the same as a real switch test.
- Leave a written "what stays on, what does not" sheet with the client. It prevents the awkward call after the first blackout.
- If multiple battery stacks are involved, verify polarity and breaker ratings at each point, including the Pylontech combiner box.
The honest price
Looking back, I should have spent one hour reading the hybrid manual's AC wiring section before placing the order. At the time, I had installed enough grid-tie inverters that I thought the backup terminal was a nice extra. It is not optional. It is the only thing that powers a building during an outage.
If you're on the B2B side—installer, integrator, or wholesaler—ask this question on your first call: "Which loads have to run when the grid is off, and what is their starting current?" If the answer is "everything," a single hybrid inverter may not be the answer. You might need a larger solution or a properly designed critical-loads subpanel. Ask before you commit.
As of January 2025, that question is now standard in our proposals. It caught the same issue on another project before we ordered the inverter. That alone paid for the mistake.
