Solar QA/QC From Factory to Commissioning: An Investor's Guide

Many of the defects that cut a solar plant’s returns are created before it exports a single kilowatt-hour: in the module factory, in the container and in the hands of installation crews. The Solar Bankability project calls these precursors: risks that originate in production, transport or installation and surface only as failures in operation ^[1]. Solar QA/QC through procurement and installation finds them while suppliers and the EPC contractor still pay to fix them. The economics are lopsided: in the same project’s model, preventive measures cost roughly EUR 1-40 per kWp once, against failure costs of about EUR 105 per kWp per year when every failure has to be fixed ^[2].

Why solar QA/QC is an investor issue

Quality assurance (QA) is the system that prevents defects: qualified suppliers, procedures, trained crews, inspection plans and audits. Quality control (QC) is the inspection and testing that finds what QA missed. An investor needs both, plus the records, because records are what survive construction.

The EU-funded Solar Bankability project analyzed failure data from 772 plants, about 442 MWp and roughly 2.4 million components ^[1]. Bad installation quality, glass breakage and PID were among the module risks with the highest economic impact. In cabling, improper installation and mixing connector types in one plant were the most prominent risks, with connection and connector failures costing about as much as the top module risks. Frequent wrong-installation tickets for inverters led the authors to recommend third-party verification of installation for all components ^[1].

The commercial structure explains why this falls on the owner. Investors want long defect warranty periods; EPC contractors want short ones and minimal extra guarantees ^[2]. Insurers try to limit their liability to failures with an external root cause, in plants that meet technical market standards and are regularly maintained ^[2]. A latent manufacturing or workmanship defect found after the defect notification period has lost its natural payer.

Having managed project delivery on more than 300 MWp of solar EPC projects in Hungary, my position is that QA/QC should be built around that gap: every check that matters should leave a dated, located and signed record before risk passes from supplier to contractor, and from contractor to owner. This article covers execution, from purchase order to final acceptance; design-stage work, including writing the QA/QC plan, is covered separately in risk-led, bankable solar park design.

Procurement: locking quality in before modules ship

Factory audits and BOM verification

Module quality varies by factory and by BOM, not only by brand. Kiwa PVEL’s 2026 PV Module Reliability Scorecard recorded at least one failure at 87% of manufacturers and 61% of BOMs, a record ^[3]. Two findings matter for execution. Delamination surged, with 45% of manufacturers showing at least one major delamination defect, driven more by materials and the lamination process than by encapsulant thickness; two BOMs from one manufacturer, with different encapsulants, both delaminated enough to cut electrical creepage distance to zero. Junction box failures fell but remain among the most consequential ^[3]. Process-driven defects like these are what production oversight exists to catch. Inspection data agree: across more than 80,000 inspections at over 340 factories, Intertek CEA rated more than 70% of factories C or D ^[10].

The BOM lock and the audit rights belong in the procurement specification, before orders are placed. In execution, the task is to prove that the factory builds what was locked: Kiwa notes that small BOM changes can alter quality dramatically and that one datasheet typically covers dozens of BOMs ^[4]. In practice:

• Run a pre-production audit a few weeks before production starts and let its findings set the scope of oversight ^[4].
• Hold the manufacturer to the contracted BOM, with no unilateral changes to specifications, manuals or price, and to the go/no-go rejection criteria and corrective-action duties in the contract ^[4].

Production oversight, flash data and pre-shipment inspection

During production, an independent inspector confirms that the specified BOM is used, that audit corrective actions are in place and that factory equipment is monitored. Batch tests, in an external laboratory or a witnessed factory laboratory, compare production modules with the BOM’s qualification results ^[4].

The Solar Bankability EPC checklist requires standard test condition (flash) data for every module, submitted to the EPC contractor for verification, recommends factory EL images of every module, and asks for the flash list to be mapped to string positions in the as-built documents ^[6]. That mapping is the baseline for any future warranty claim.

Before shipment, a random sample is inspected against the specification, and a container loading inspection confirms that the monitored modules are the ones shipped ^[4]. Sample sizes usually come from ISO 2859-1, whose AQL-indexed plans use the threat of lot rejection to keep the producer’s process average at least as good as the agreed acceptance quality limit, while capping the buyer’s risk of accepting an occasional poor lot ^[11]. They do not certify every module, so the contract must fix the inspection level, the AQL for each defect class and the consequences of a rejected lot.

Inverters, transformers, structures and cables

For inverters, transformers and switchgear, factory acceptance tests (FAT) take the place of production oversight. FAT reports, cable type and routine test reports and tracker commissioning reports all belong in the handover file ^[6]. I recommend witnessing FAT for long-lead items: a defect found in the factory costs a retest, while the same defect found on site costs a replacement shipment and a schedule slip. Every major component should be visually inspected on delivery ^[6], and galvanized steel checked for coating thickness.

Transport, handling and incoming inspection

Solar Bankability found that module condition on delivery is often unclear and the origin of damage untraceable, because outgoing quality is undocumented and packaging and handling requirements are not properly specified or followed ^[1].

Dhimish and co-authors tracked 186 modules from packaging to installation. None had cracks before packing; after standard shipping, site EL found cracks in 2.2% of them, with estimated power losses of 0.53% to 1.43% depending on crack size. Six months after installation, every cracked module had developed hotspots, with temperatures raised by 10 to 20 °C, and PID affected the cracked modules more than a crack-free reference ^[5]. The sample is small, but the chain from handling damage to hotspots is clear. In Clean Energy Associates’ EL testing of more than 300,000 modules at 150 sites, complex cracks, which arise from handling, installation or severe weather, appeared at 76% of sites, and only EL reveals them ^[12].

Incoming QA needs three things:

1. Requirements up front: transport and handling requirements in the contracts, and a handling protocol with training for field workers ^[6].
2. A baseline at each handover: arrival condition recorded, damaged pallets quarantined, and a module sample tested after delivery and before acceptance by an accredited independent laboratory, against criteria written into the EPC contract ^[6].
3. A way to date damage: UV fluorescence imaging can establish when cell cracks formed, separating installation cracks from hail damage, and can reveal mixed encapsulant or backsheet types within a module population ^[9].

Without a baseline at each transfer of custody, a cracked module becomes nobody’s fault and, by default, the owner’s cost.

Installation QA: hold points, torque and connectors

Installation quality is governed by the inspection and test plan (ITP): for each activity, the specification, acceptance criterion, inspection frequency, record and inspector. Hold points stop work until an inspection is signed off. My recommendation is to place them wherever later work hides or locks in earlier work: piles before structure assembly, structures before module mounting, DC strings and connectors before energization, and MV terminations before energization. Piles should be accepted against the structure supplier’s tolerances for position, embedment, plumb and head height, with an as-built survey, while correction is still cheap.

Solar Bankability’s list of common installation failures reads like a site checklist: broken modules, loose module clamps, wrongly installed or configured inverters, incomplete structures, loose or unprotected cables, missing equipotential bonding, mixed grounding materials and badly installed monitoring sensors, mostly caused by lack of know-how, short installation times and poor planning ^[1]. Free-hanging cables, cables exposed to sunlight and broken cable ties belong on the same list ^[1].

Connectors and terminals: where workmanship becomes fire risk

A Fraunhofer ISE and TÜV Rheinland-led analysis of about 180 fire and heat-damage cases caused by PV systems found the DC side dominant and field-made connections particularly sensitive ^[13]. Where a cause was identified, installation faults and product defects each accounted for roughly 35% of cases. Workmanship failures included connectors not fully plugged or poorly crimped, unfastened screw terminals, and cross-mated connectors from different manufacturers that overheated hundreds of contact pairs in one large plant. The authors also noted that screw terminals need controlled torque to the maker’s specification ^[13].

Cross-mating persists because it passes commissioning: a mismatched pair usually carries current without visible problems, then degrades into losses, burnt connectors or arcing ^[1]. A Bern University of Applied Sciences study of connectors taken from operating plants found that only cross-connected pairs exceeded the 5 mΩ contact-resistance limit that IEC 62852 sets for new connectors; one exceeded 45 mΩ and was too deformed to separate. The sample was small, but the result supports manufacturers’ warnings ^[14].

The controls belong in the ITP:

• Field connectors of the same manufacturer and type as the module connectors, crimped with the right tools ^[13].
• Trained, authorized crimpers, tool calibration records and sample crimp tests.
• A visual check of every mated pair, with mixed pairs rejected outright.
• Calibrated torque tools, recorded values and marked fasteners.

Commissioning and acceptance: proving the plant before handover

IEC 62446-1 sets commissioning tests, inspection and documentation for grid-connected systems, with test regimes scaled to system size and complexity ^[7]. Category 1 covers earthing and bonding continuity, polarity, string open-circuit voltage and current, functional tests and insulation resistance. Category 2 adds string I-V curves and infrared inspection, with wet insulation testing available as an additional test ^[7]. For utility-scale plants I treat Category 2 on every string as the minimum: I-V curves are the cheapest way to find damaged modules, wrong module counts and high-resistance connections before the contractor demobilizes. Thermography should follow IEC TS 62446-3, which sets equipment, conditions, procedure and personnel requirements and classifies thermal abnormalities ^[8].

The IEA PVPS Task 13 report on mobile test equipment adds three cautions ^[9]:

• Inspecting every module of a multi-megawatt plant is not feasible, and sampling methods are not internationally harmonized, so the EL and module-test sampling plan must be agreed in the contract.
• EL shows cracks but not power, so imaging should be paired with power measurements.
• String-level tests give only average module performance, which may not support a warranty claim that needs proof for an individual module.

Acceptance should be two-staged: a short-term performance test at provisional acceptance, with defined duration, irradiance threshold, monitoring set-up and temperature correction, plus a visual inspection (with IR as best practice) and as-built documents to IEC 62446; then final acceptance two years later, accounting for degradation and availability ^[6].

What QA/QC costs compared with what defects cost

Solar Bankability is one of the few public sources that prices both sides. Its 2016 cost ranges were ^[2]:

• EL imaging, IR inspection and PID testing: EUR 0.5-1 per kW each.
• Module component testing and EPC qualification: EUR 1-10 per kWp each.
• Design review plus construction monitoring: EUR 10-40 per kWp.

Against these, the summed cost priority number of the main failures, with every failure detected and fixed, was modeled at about EUR 105 per kWp per year. Qualifying the EPC contractor alone cut it to about EUR 75, adding design review to about EUR 40, and the best combinations of preventive measures saved up to about EUR 90 per kWp per year. In 99% of the combinations tested, mitigation paid off, and prevention outperformed earlier detection ^[2].

An IEA PVPS Task 13 case study reaches the same conclusion: for a 10 MWp plant affected by PID, 20-year profit was 48% below expectations without mitigation, but only 5-6% below with mitigation, including its cost ^[15].

These are model results from one failure database at 2016 prices, best read as a ranking rather than a forecast for a specific plant. The direction held across all three cost scenarios tested ^[2]: preventive QA is a one-off cost per kWp, while the defects it prevents cost money every year.

A QA/QC roadmap for investors

The roadmap starts once the supply contracts are signed and draws on Kiwa’s procurement practices, the Solar Bankability EPC checklist and the IEC test standards ^{[4, 6, 7, 8, 11]}. The sign-off column is my recommendation and should follow the contract structure.

Questions to ask your EPC contractor

1. Which exact BOM is in the module supply contract, and which independent test data cover it?
2. Who audits the factory and oversees production, and are they independent of the manufacturer?
3. What sampling plan (standard, inspection level, AQL per defect class) applies before shipment and on delivery, and what happens to a rejected lot?
4. Will we receive flash data and factory EL images by serial number, and the as-built module-to-string map?
5. Which ITP hold points exist, and who can release them?
6. How are DC connectors controlled: matching make and type, approved crimp tools, trained crimpers?
7. Which IEC 62446-1 test category applies, to how many strings, and who performs the IR and EL sampling?
8. How are nonconformances closed, and how is closure linked to payments and acceptance?

Conclusion

QA/QC in procurement and installation is not a line to negotiate down at the end of a tender. It decides whether a defect is found while a manufacturer or contractor still pays for it, or after it has become the owner’s lost yield, warranty dispute or fire risk. The tools are standard and the costs are modest. What investors must add is ownership: approve the plan, attend the hold points through an independent engineer, and accept the plant on evidence.