-
Who this checklist is for
-
Step 1: Lock the system architecture before the PO goes out
-
Step 2: Match the battery datasheet to the duty cycle
-
Step 3: Read the inverter charger efficiency curve, not the nameplate
-
Step 4: Treat the module and mounting as one system
-
Step 5: Match storage to the inverter's charging behavior
-
Step 6: Commission against metered data—including the smart meter
-
Step 7: Freeze the documentation before the site crew opens the boxes
-
Common mistakes I see on first submissions
Who this checklist is for
I'm the quality and brand compliance manager at a renewable energy integrator. My job is to review every component package before it reaches a job site—roughly 180 packages a year. In 2024 I rejected 22% of first submissions. Not because vendors are incompetent. Because specs drift, documentation goes stale, and 'same factory' rarely means 'same spec'.
This checklist is for installers, system integrators, and commercial buyers about to send a PO for an off-grid or mobile energy system. Seven steps. In order. It won't fix a bad system architecture, but it'll catch the mismatches that turn commissioning into a fire drill.
Step 1: Lock the system architecture before the PO goes out
DC bus voltage. Load profile. Autonomy target. Grid-interactive or island. Write these down before you compare a single quote.
The trap here is jumping straight to component pricing. I've watched three projects in the last two years where the lowest per-component quotes produced the most expensive systems, because nobody checked the interfaces. A 24V bank paired with a 48V inverter charger. Bifacial modules mounted flat on a low-albedo roof where the back-side gain never showed up. People think cheaper components make a cheaper system. The causation runs the other way: integration cost drives total cost, and component price is a rounding error next to a redo.
Checkpoint: one page, signed off by the person who'll commission it. If the signer and the specifier aren't the same person, get both signatures.
Step 2: Match the battery datasheet to the duty cycle
For marine work this usually means a Victron marine battery, and the datasheet needs reading, not skimming. Cycle life numbers are meaningless without the test conditions attached: depth of discharge, C-rate, ambient temperature. A battery rated for 3,000 cycles at 80% DoD and 25°C won't deliver that at 45°C in an engine room locker.
What I check on every marine battery submission: continuous and peak discharge, BMS communication protocol (CAN, VE.Can, Modbus), and evidence of salt-spray and vibration testing to ABYC E-13 or IEC 62619. If the vendor's only environmental evidence is a marketing PDF, that's a rejection.
Here's the one that still stings. In Q2 2024 I skipped the incoming terminal-torque check on one pallet. Same factory, same lot range as the previous shipment—what were the odds? The odds caught up. Nine units were below the manufacturer's terminal torque spec. We caught it at install, but two boats slipped a week and the yard charged us standby time. Now the torque wrench comes out on every pallet. No exceptions.
Step 3: Read the inverter charger efficiency curve, not the nameplate
Victron Energy inverter chargers usually get specified on the headline efficiency number. That number is the peak, and the peak lives in a narrow band around 30–50% load. Commercial off-grid loads spend a lot of time at 10–20%, where the curve is less flattering.
Ask for the EU weighted efficiency or a full load-versus-efficiency table. Then check three things the datasheet doesn't headline: standby draw in search mode, charger programmability against your battery's absorption and float voltages, and parallel or three-phase capability if the site will grow. Safety certification should be visible on the document—IEC 62109-1 and -2 for the inverter side.
If the efficiency table only has three points, ask for the full curve. If they can't send it, that tells you something about the support you'll get at commissioning.
Step 4: Treat the module and mounting as one system
Aptos 550W bifacial solar panels illustrate the problem well. The bifaciality factor on the datasheet is measured under IEC 60904-1-2 lab conditions. On a tilted roof at 15 cm mounting height with low-albedo shingles, the back-side gain is a fraction of the headline number. On a white membrane roof at 1 m height, it's real.
What I need before signing off a bifacial array: the measured bifaciality factor, the temperature coefficient of Pmax, mechanical load test evidence to IEC 61215, and—critically—the MPPT input window of whatever controller is downstream. A Victron SmartSolar MPPT has a maximum PV input voltage that has to clear the module's Voc corrected for the coldest expected site temperature. That's basic. It still gets missed.
People assume bifacial always beats monofacial on yield. Actually, on a low-albedo, low-height install, a monofacial module can win on cost per kWh delivered. The assumption is the relationship. The reality is the mounting condition.
Step 5: Match storage to the inverter's charging behavior
Fronius battery storage gets specified on kWh and backup switchover time. Both matter. What gets missed is the handshake between the storage system and the inverter charger's charge algorithm. If the battery wants a three-stage charge and the charger is set to a fixed float, the battery will live a short, unhappy life.
I ask for the charge curve table, the SOC recalibration interval, and evidence of UL 9540 or UL 9540A listing for the complete system—not just the cells. A cell certificate is not a system certificate.
Step 6: Commission against metered data—including the smart meter
Every system I sign off gets a metered cross-check. If the site has a utility AMI meter or a revenue-grade smart meter, know how to take a meter reading from a smart meter before you start. On most AMI displays you're cycling through registers with a button; on commercial revenue meters there's usually a scrolling display with a kWh register and a demand register. Write down the reading, run a known load, take a second reading, divide the delta by the time. That's your real power. Compare it to what the inverter charger or monitoring portal reports.
A 5% gap is normal. Ten percent means a CT or shunt problem. A 30% gap means someone wired the shunt downstream of the charger and the system has been logging fiction since day one. I've seen that on a small commercial job where we skipped the cross-check because 'the portal looked fine'. Six weeks later the utility bill didn't match the commissioning report. The cross-check takes twenty minutes. The correction took a return visit and a difficult conversation.
Step 7: Freeze the documentation before the site crew opens the boxes
Model number, firmware version, serial range, and a written firmware-update policy. That's the package. If two identical-looking units arrive with different firmware, they're not identical units, and the parallel configuration may not behave the way the manual describes.
Also worth doing: photograph the labels on the actual delivered units before installation. When a warranty claim comes in eighteen months later, that photo is the difference between a straightforward replacement and a circular argument about what was originally supplied.
Common mistakes I see on first submissions
- Vendor self-test reports passed off as third-party evidence. A manufacturer's internal test is useful. It isn't the same as an IEC or UL listing.
- Bifacial gain calculated at the datasheet factor. Halve it on low-albedo or low-height installs until you have site data.
- Smart meter readings used as a substitute for load verification. They're a cross-check, not a replacement for a clamp test or load bank.
- SOC calibration deferred until an alarm fires. By then the battery's already been cycling in a bad state.
- Firmware left floating. Pin it before commissioning, or you'll be debugging a moving target.
One last thing. The lowest quote has cost us more in roughly 60% of the projects I've reviewed over four years. Not because cheap components fail more often—sometimes they do, sometimes they don't. Because the cheap quote usually excludes the integration work, the documentation package, and the commissioning support that make the difference between a system that runs and a system that gets signed off. Count the total. Then decide.