Insulation resistance, string IV curves, earth continuity and the anti-islanding trip, what a pass and a failure look like on each, and the documents that have to be in the handover pack before anyone signs.
Handover day usually goes the same way. A folder appears on the table, somebody says the plant has been fully tested, and a form gets signed. Six months later a string is found to have been dead since day one, and the only record of what was tested is that signature.
Commissioning is where promises become records. The tests are not difficult and the instruments are not exotic. What matters is that they are done, that the results are written against a position on a drawing, and that the client or their consultant stands there for the ones that count.
IEC 62446-1 sets out the minimum documentation, the commissioning tests and the inspection regime for a grid-connected photovoltaic system. It separates a basic category of tests from an additional category that includes string IV curve tracing and infrared thermography. Naming it in a specification changes the conversation, because it turns a subjective argument about thoroughness into a checkable requirement. On anything of industrial size, specify the additional category.
There is also a practical division that decides which tests you should insist on attending. The dead tests are done before the array is energised, and once the plant is live they cannot be repeated without shutting it down and isolating sections. The live tests can be repeated later at some inconvenience. Prioritise your attendance accordingly.
On insulation resistance, one practitioner point is worth more than the rest of this section. Test it when the roof is wet, or at least test it again after rain before you accept the plant. A great many DC insulation faults only appear when water is present, which means a plant that passes on a dry February afternoon can nuisance trip its way through the entire monsoon. If the contractor will only test on a dry day, ask for a repeat test written into the defects liability period.
String IV curve tracing. Worth paying for on any industrial plant, because the shape of the curve tells you things a voltage reading cannot. The curves must be translated to standard test conditions using measured plane-of-array irradiance and measured module temperature, so an irradiance sensor and a back-of-module temperature probe are part of the test kit. An IV report with no measured irradiance is a set of numbers with nothing to compare them to.
| What the curve shows | What it usually means |
|---|---|
| A step or notch in the curve | A bypass diode is conducting: shading, a cracked cell, or a failed diode |
| Reduced current, correct voltage | Soiling, uniform shading, or module mismatch in the string |
| Reduced voltage, correct current | One or more modules effectively out of circuit |
| Rounded knee, sloping toward Voc | Series resistance: a poor connector, an undersized or overlong cable, a loose terminal |
| Sloping plateau near Isc | Shunt path or cell level defect |
Infrared thermography. Modules at a decent irradiance, and separately the terminations inside every DC and AC enclosure under load. Hot cells and hot connectors on the roof are useful. Hot terminations in the AC panel are the actual fire risk on most plants, and they are found in ten minutes with a camera and never found at all without one.
Inverter parameter records. Not the name of a country code setting. The actual values written into each inverter, printed or screenshotted from the unit itself: under and over voltage stages with their trip times, under and over frequency stages with their trip times, reconnection delay, ramp rate, the reactive power or power factor mode, and any export limit. The utility granted approval against a set of settings, and if a technician changes one during a fault-finding visit two years from now, this record is the only way anyone will know.
Anti-islanding trip test. Witness this one personally. With the plant generating into local load, the incoming supply is opened at the agreed point and the inverters must cease to energise the network within the required time. Record the measured time, not a tick in a box. This is the test that protects a utility lineman working on a supposedly dead feeder, and it is also the test most likely to be certified without being performed, because performing it requires a planned outage that is inconvenient for everybody.
Alarm and monitoring verification. Force an earth fault alarm and an inverter fault and confirm each one arrives where it is supposed to arrive: the portal, the email address, the phone. An alarm that only appears on the inverter's own screen, in a locked room, is not an alarm.
A recorded baseline. Measured AC output against measured plane-of-array irradiance and cell temperature, over a clear period on the day. This is not a full performance ratio test, which needs weeks. It is the anchor measurement you will compare against for the next twenty years, and it exists only if somebody writes it down on the day.
Copy this into the tender. Retention against its delivery is the only thing that reliably produces it.
A failed test at commissioning is a good outcome. It means the process is working. What matters is what happens next, and the answer is that it goes on a snag list with a named owner and a date, and the takeover certificate is issued subject to it, with retention held against closure. Do not sign a clean certificate on a promise to come back, because the commercial leverage you have on the day of handover is the most you will ever have.
And be reasonable about the ones that are not the contractor's fault. If the anti-islanding test cannot be done because the utility will not grant an outage that week, record the reason, hold the retention, and schedule it. The point of the discipline is a plant whose condition is known, not a stack of paper.