BS EN 62446-1 in practice: what auditors actually look for
We've signed off a lot of 62446-1 reports and watched a few get pulled apart on audit. Here's the bit between the standard and the clipboard that nobody writes down.
BS EN 62446-1 is a 60-page document. The forms most installers use are six pages. The gap between the two is where every failed audit lives.
We’ve signed off on a lot of 62446-1 reports over the years — for our own installs, for inherited fleets we took over, and for one or two cases where an owner had asked us to come in and check the previous installer’s work. Some of those reports got audited. A handful failed the audit. We learned more from the failures than from the standard itself.
Here’s what the people pulling these reports apart in 2026 actually look at, and what they care less about than the field engineer usually thinks.
What the standard says, briefly
BS EN 62446-1:2016+A1:2018 covers documentation, commissioning tests and periodic verification for grid-connected PV systems. For a UK commercial install it’s effectively mandatory — MCS certification, IEC 61730 module compliance and most insurance products all reference it. For periodic inspection it’s the obvious benchmark even when the contract doesn’t name it explicitly.
The standard splits into three buckets:
- System documentation — single-line diagram, string layout, module and inverter datasheets, AC isolator locations, labels.
- Verification tests — visual inspection, continuity of protective earthing, polarity, Voc per string, Isc per string, insulation resistance, AC functional checks.
- Reporting — the actual signed-off document with measured values, photos, and a statement of compliance.
On paper, simple. In a roof-edge harness in February, less so.
What auditors do first
Almost every audit we’ve watched starts with the same two checks before the auditor even reads the test values:
- Photo provenance. Are the photos clearly from this site, on this date, showing the equipment described in the report? GPS-tagged photos make this trivial. Phone screenshots of last year’s site dropped into Word make this impossible to verify, and that alone has been enough to get a report flagged as “non-compliant pending re-issue”.
- String labelling continuity. Does the string referred to as “Inverter 1, MPPT 2, String B” on the test sheet match the label on the DC combiner, match the label on the inverter terminal, and match the single-line diagram? Three places. Inconsistency between any two of them is a finding.
If those two checks fail, the auditor’s mindset is set before they read a single test value. The rest of the report is being read for problems, not for confirmation.
The verification tests, ranked by how often they get flagged
In our experience, the test values that come back queried (or that quietly fail an audit when nobody pushes back) are, in rough order:
1. Insulation resistance
Required: > 1 MΩ at the test voltage specified for the system voltage class. For a typical 1000 V DC commercial array that’s a 1000 V test, and “comfortable” readings are in the hundreds of MΩ.
What gets flagged: readings of ”> 999 MΩ” with no actual figure, or readings clearly clipped at the meter’s display ceiling. A reading of >999 MΩ is technically a pass but tells nobody anything about the next inspection’s trend. Record the actual value. A megger that won’t go above 999 needs replacing or noting in the test conditions.
What also gets flagged: readings taken at a lower test voltage than the system class requires. A 500 V test on a 1000 V system is a fail dressed up as a pass.
2. String Voc within tolerance
Required: each string’s open-circuit voltage within typically ±5% of expected, temperature-corrected. The expected value is the module Voc × number of modules, corrected for cell temperature using the module’s voltage coefficient.
What gets flagged: Voc readings that are perfectly consistent across all strings (suggesting they were calculated, not measured), readings without a recorded ambient temperature, or readings that fall outside tolerance with no explanation.
The temperature correction is the bit most engineers fudge. On a cold February morning the actual Voc is meaningfully higher than the STC value, and we’ve seen perfectly healthy strings flagged because the engineer didn’t apply the coefficient.
3. String Isc within tolerance
Required: each string’s short-circuit current within typically ±5% of the others on the same MPPT, irradiance-corrected.
What gets flagged: Isc measurements with no recorded irradiance figure. A handheld irradiance meter reading goes on the test sheet, or the test is done at a time of day where shading is a confounder and that’s noted explicitly. “It was a sunny day” is not a recorded irradiance.
This is the test that links most directly to operational fault detection. The same string-to-string comparison logic that catches a 30% current drop in O&M is what catches a wonky combiner at commissioning. We wrote about string-level fault detection in more detail; the field test version of the same idea.
4. Continuity of protective earthing
Required: a measured low-resistance continuity test from every accessible exposed metal part to the main earth, with a recorded value.
What gets flagged: a tick-box “Pass” with no value. The standard wants a number. < 1 Ω is the usual benchmark; values close to or above 1 Ω need explanation (length of run, conductor size).
5. Polarity
Rarely flagged — it’s binary. But every now and then a report comes back with a noted reverse-polarity finding and no record of what was done to correct it before sign-off. If you find a reversed string, the report needs the corrective action recorded, not just the finding.
The documentation traps
Once the test values are read, the auditor moves to documentation. Three places where reports lose points:
- Single-line diagrams that don’t match the install. The classic: a contractor changes the array layout at install time (a roof obstruction, a re-roofed section), and the SLD in the report is still the design version. Update the SLD or annotate the deviations on it.
- Missing module / inverter / cable datasheets. All three are required as appendices. “Available on request” is not compliance.
- Labels. The standard is prescriptive about AC/DC isolator labels, warning signs, and emergency disconnect labels. Photographing them as part of the visual inspection is the easy fix. We’ve watched reports get re-issued because the inverter was missing the dual-supply warning label.
What auditors care less about
Counter-intuitively, perfect typography and a glossy PDF are not what saves a report. A scanned hand-written sheet of paper with measured values, dated photographs and a clear signature has passed audits we’ve seen fail glossy ones. Substance over polish.
The auditor wants to know: did this engineer actually go to this site, on this day, and measure these values with these instruments? A report that answers that question convincingly is fine. A report that obscures it is not.
How we ended up rebuilding the form
After enough of these reviews, we got annoyed with the gap between what the standard requires and what the typical PDF form makes easy.
So the 62446-1 module in SolarFleet takes the testing flow on mobile. Every reading is timestamped. Every photo is GPS-tagged and date-stamped. Voc readings prompt for an ambient temperature; the coefficient correction is applied on the spot. Isc readings prompt for an irradiance figure. Insulation readings refuse to accept “>999” — the engineer either records the actual value or explicitly notes the meter ceiling.
The output PDF is no glossier than anyone else’s. But every value on it is one the auditor can trace back to the moment it was measured. That’s the only thing the audit is actually checking.
The bit nobody tells the new engineer
Commissioning a system to BS EN 62446-1 the first time is genuinely difficult — there’s a lot to remember and most of it is unforgiving in the field. We’ve all signed reports we’d write differently a year later.
What changes things is treating each report as a baseline for the next inspection. A periodic verification two years later is mostly a comparison exercise against the commissioning values: has the insulation resistance dropped? Has any string drifted in Voc or Isc? That comparison only works if the original numbers are honest. Padded commissioning readings make periodic inspections useless.
Record what you measured. Note what you couldn’t. Sign your name. Move on.
SolarFleet ships the 62446-1 inspection flow with every site, on every plan. Test values, photos and signatures sync from the field. Start free with 2 sites or book a walkthrough.
Josh runs InspireGreen, a solar installer based in Cardiff, and builds SolarFleet — the O&M platform we use to monitor our own sites. Most posts here come straight out of the work: a case we dealt with, a feature we shipped, or a thing we wish we'd known earlier.