Project desk: +1 888 482 2385 | [email protected] Global EPC support | EN

Storage insight

Why Off-Grid Solar Systems Fail: A Quality Manager's Field Notes

2026-11-03 / Elias Bergstrom

The 9 PM Call

A system that was commissioned three weeks ago is producing 40% of its expected output. The installer is sure it's the MPPT controller. The customer is sure it's the panels. Somebody on a Facebook group says it's the battery's BMS. Everyone has a theory, and none of them has pulled a data log.

I've been on the receiving end of that call more times than I can count. I handle quality and brand compliance for a solar equipment distributor—I review every shipment of charge controllers, inverters, and LiFePO4 batteries before it reaches our installers. Roughly 200 controller units a month, plus whatever drifts through the audit queue. In our Q3 2024 review cycle, we rejected 11% of first deliveries for spec drift, firmware mismatches, or documentation that didn't match what actually shipped (this was after we tightened the verification protocol in 2022, and it was still 11%).

Here's what I've learned from those rejections: the hardware is almost never the reason a system underperforms in the field. When a freshly installed system fails to deliver, the failure started in the spec sheet—not on the production line, and not with the panel that happened to be in partial shade.

What People Think Is Wrong

Installers, and honestly most distributors, tend to diagnose off-grid problems in a predictable order: controller, then battery, then panels. It's a natural sequence—those are the three components that show up in the fault log, so those are the three things people stare at.

Nine times out of ten, when we pull the controller log, the controller was doing exactly what it was configured to do. It was charging at the current limit someone set. It was switching to float at a voltage someone typed in. It was balancing a battery bank that was never properly sized for the load. The MPPT algorithm was fine. The spec wasn't.

I want to walk through three of the most common culprits, because they're not what the forum threads say they are—and one of them is genuinely counterintuitive.

Solar panel shade is not a controller problem

The default assumption when a system underproduces in a partly shaded location is that a better controller will save it. Buy a premium MPPT, tilt the panels more aggressively, and the shade losses disappear.

That's backwards. People think a better controller solves shade problems. Actually, shade losses are string-design losses—the controller can only minimize what physics has already taken. A shaded cell drops the current of its entire series string. Bypass diodes cap that loss, but the string still falls to the current of its weakest cell in the worst case. The MPPT can't harvest electrons that the string can't pass. If the string design ignores the shade profile of the site, you've already lost the fight before the controller does anything.

What a good MPPT—and I'll include the EPEVER MPPT controllers we distribute in that group, because their tracking under partial shade is genuinely competitive—does is harvest the remaining mismatch efficiently. It bridges a gap. It doesn't close a hole.

The installers on our account list who get this right do one thing differently: they walk the site at noon and at 4 PM before specifying a string layout. Not because it's clever. Because it's the only honest way to see where the shadow of the chimney lands in February versus July.

The controller and the battery speak different languages

The second misdiagnosis is more subtle. A system with LiFePO4 batteries starts cutting off early, or floats too high, or never reaches full charge. The controller gets suspected immediately.

Most of the time, what's actually happening is that the controller is running a lead-acid default profile that nobody changed. The whole flexibility conversation around lithium settings—the EPEVER solar charge controller lineup is decent here, with dedicated lithium profiles and user-adjustable parameters—only helps if someone opens the menu and configures it. I've seen 48V systems running a 12V default because the installer assumed the auto-detect handled it. That's not a hardware fault. That's a five-minute setting that didn't happen.

This is where I'd tell you to check your vendor's documentation before blaming them, and I mean it for two reasons: because their defaults are (usually) reasonable once you read them, and because when a genuine spec drift does happen, you'll actually know the difference.

The causation runs the other way on component quality

Here's the one I keep coming back to, because it shapes how our team talks to customers. People assume expensive systems are more reliable because they use more expensive components. The parts are better, so the system holds up. Fine.

But in the audits I run, that's not what the data shows. The systems that hold up are the ones that were specified correctly. Correctly specified systems tend to cost more, because proper sizing, proper string design, and proper battery settings mean buying the right hardware instead of the cheapest hardware that technically meets the nameplate. The causation runs the other way. The reliability isn't caused by the price tag. The price tag is a side effect of doing the process right.

When a customer asks how much a Tesla solar and Powerwall setup costs before telling me their load profile, I understand the impulse—everybody wants a number—but the number that matters comes after the spec. A system sized wrong at any price fails. A system sized right at a mid-range price outperforms it every time.

What Getting It Wrong Actually Costs

I've mentioned what we reject on the incoming side. The cost on the outgoing side is worse.

In our 2023 audit, we found that a 40-unit rollout to a commercial off-grid client was running controllers with firmware two versions behind the current release. Nothing broken, per se—but the older firmware had a float hysteresis behavior that, combined with that client's specific battery bank, was causing the bank to cycle in a way nobody specified. Fixing it meant pulling 40 units, flashing new firmware, and re-commissioning the sites. That firmware issue cost us a $4,200 redo and delayed the client's opening by eleven days.

Eleven days on a commercial install is not a small thing. Retail opening dates slip. Contractors get paid late. Word travels.

For contrast, the process change we made after that audit—verifying firmware version against the spec sheet on receipt, before units go anywhere—cut our field-failure rate on controllers from roughly 4% to under 1% within two quarters. Same vendor, same products, same price. Different spec discipline.

And yes, there's something genuinely satisfying about watching a system come online at exactly the numbers the spec predicted. Not "close enough"—exactly. After six weeks of vendor back-and-forth on firmware and one renegotiated contract, seeing a commissioning report hit 97.2% of design output is the payoff. It's not glamorous. But it's the job.

What Actually Fixes This

I'm not going to give you a 12-step framework. The fix is shorter than that, and the reason it's short is that most of what's wrong here is procedural, not technical.

Three things, in order:

  1. Specify against the site, not against the panel nameplate. Walk it. Note the shade at two times of day. Match the string to the shade, not to a spreadsheet.
  2. Verify the configuration menu against the spec sheet on receipt. Not on commission. On receipt. If your lithium profile is set to lead-acid defaults, you want to know that before the unit leaves the warehouse, not after.
  3. Export the log after commissioning, and compare it to the design output in the first 72 hours. If the delta is large, you're looking at a spec problem, not a hardware problem—and the first 72 hours is the cheapest time to find out.

The same pattern shows up in adjacent categories, by the way. We've helped a couple of installers work through EV charger safety testing protocols, and the finding was identical: the equipment passed every bench test, and then failed at the site because the installation spec hadn't accounted for the actual electrical environment. It's not a coincidence. It's the same failure mode—component-level correctness applied to a system-level problem.

Bottom line: if you're chasing a system failure right now, the controller is probably fine. The battery is probably fine. The question worth asking yourself is whether anyone wrote down, in a single readable document, what the configuration was supposed to be before the install started. If the answer is no, start there.

Elias Bergstrom

Elias Bergstrom

Elias Bergstrom is a solar array and balance-of-system analyst specializing in mounting structures, trackers, DC cables, connectors, isolators, fuses, combiner boxes, grounding, and surge protection. He applies IEC 62548-1 design requirements while checking cold-condition open-circuit voltage, corrected short-circuit current, conductor ampacity, voltage drop, protection coordination, fastener torque, bonding continuity, and structural interface loads. His application notes help designers and installers coordinate safe array architecture, compatible components, roof or ground conditions, inspection access, and commissioning documentation.

Ask a Catl storage specialist