Table of Contents

Solar inverter problems show up as a change in behavior: no AC output while the array is in sun, output that falls on hot afternoons, a trip that returns as soon as the grid returns, or a fan that runs and the cabinet that still overheats. The useful first step is to separate a grid event, a thermal limit, and a failed part. Those three have different remedies. Replacing the inverter is the remedy for only one of them.

Pacific Northwest National Laboratory tells you how not to start. Modules produce in daylight even if the inverter is off and even in cloud. The emergency order is the main AC disconnect, then the main DC disconnect, and only trained people should connect or disconnect wiring. A problem-solving guide that begins with “unplug the DC connectors” is the hazard, not the fix. What follows is how to read the symptom. It is not a procedure for opening a live cabinet.

A trip is often the grid, not a broken power stage

IEEE Std 1547-2018 requires a distributed energy resource to cease energizing an unintentional island within 2 seconds. The National Renewable Energy Laboratory primer describes that as a built-in, type-tested function on most residential and commercial units in the US path, typically under UL 1741, and it has to keep working when ride-through and voltage support are enabled. So an inverter that goes quiet when the utility goes quiet can be doing the job. The symptom “zero watts during an outage” is, for an ordinary grid-tied unit, the specified behavior. The solution is not a reset. It is understanding that backup requires different equipment, which this inverter was not asked to be.

Nuisance trips are the harder case. Ride-through is supposed to keep the unit online through disturbances the bulk system wants it to survive, and tripping still outranks ride-through when the voltage or frequency is actually out of bounds. If the utility’s required setting file was never loaded, the unit can trip on events it was supposed to ride through, or ride through events it was supposed to leave. The National Laboratory of the Rockies found that ride-through can slow island detection and that, in the cases tested, detection still finished inside 2 seconds. That research does not diagnose your alarm. It does say the settings and the listing have to match before anyone condemns the hardware. Ask for the last setting file and the event log. “Fault” with no quantity and no time is not a diagnosis. A log that names the quantity is the point of a solar monitoring system.

Heat looks like a power problem

A Sandia and NREL thermal study measured inverters in the field and found component temperatures rising with the power the unit was dissipating, following the day and the season, and changing with wind. An inverter that produces well at 10 in the morning and folds back at 2 in the afternoon, on a clear day, is often reporting that thermal path, not a failed string. The fold-back threshold is in the manual for that model. Copying another model’s number will either create a false alarm or hide a real one.

Cooling parts are a known wear item, not a mystery. A 2013 workshop record says fans and blowers need periodic replacement, and that liquid cooling was viewed as practical mainly on large utility inverters. A later fan-reliability study found humidity, not just temperature, dominating modeled life, with a hot-humid climate much harder on fans than a hot-dry one, and with fans inside the enclosure lasting longer in the model than fans outside it. If the intake is blocked, or the filter PNNL says to clean on the manufacturer’s schedule is packed with dust, the thermal study’s temperature rise happens earlier in the day. Cleaning a filter and replacing a fan are maintenance. They are still electrical-equipment work. The person who does them needs the shutdown procedure, not a vacuum and the cover off while the DC side is live. Filter and fan care is treated on its own in cooling-fan maintenance.

Some “inverter faults” are array faults the inverter is reporting

PNNL’s production check is the separator. Compare output with the estimate for that month and with the same month last year. The page notes that sunlight itself can sit about 10 percent off the average, and that production outside the expected band is when troubleshooting starts. Low output can be a soiled array, which PNNL calls the largest operations issue and which is often hard to see, or it can be a dead string, a blown fuse, or a tracker stuck because one input is shaded. Washing on a guess, or swapping the inverter on a guess, both spend money on the wrong object.

A string that shares one tracker with a shaded neighbor will pull the healthy modules down. That can look like “the inverter is weak” when the index that is wrong is the tracker assignment. More on what those indexes mean, and what they do not, is in solar inverter performance indexes. The fix is to identify the input, not to raise a power-factor setting and hope.

What to do with the symptom in front of you

What you see Check before replacing anything Who acts
Zero output during a utility outage Whether this unit is grid-tied and required to stop. Under IEEE 1547-2018, ceasing to energize is the specified response. No repair. Backup is a different design.
Trips on days the grid seems normal The event log and whether the utility setting file is loaded. Ride-through and trip are both legitimate, in an order. Qualified person, with the utility’s settings, not a factory reset as a habit.
Falls off on hot afternoons Derating curve, airflow, filter, and whether the room or the sun on the cabinet matches the thermal study’s drivers: load, season, wind. Filter cleaning per the manual. Fan or power-stage replacement only after shutdown by someone qualified.
Low all day, not just when hot Month versus expected production, soiling, a single weak string, blown fuse. PNNL’s band for “sun was just low” is about 10 percent, not 40. Inspection first. Electrical repair stays with the authorized maintainer.

PNNL also says to keep the log. A one-time trip that was a grid event becomes a warranty claim or a repeated thermal fault only if somebody wrote down the time. Resetting the alarm and walking away deletes the only evidence that distinguishes those rows.

Frequently asked questions

Should a solar inverter be restarted every time it faults?

A remote reset can clear a ride-through or a grid event that has already ended. It does not clean a filter, replace a failed fan, or repair a string. If the same alarm returns, the log and the setting file come before another restart. Opening the unit is qualified work.

Does zero output mean the inverter has failed?

Not during a utility outage, if the unit is grid-tied. IEEE 1547-2018 requires it to stop energizing an island. Failure is the better description when the grid is present, the array is lit, airflow is normal, and the unit still will not run after the conditions in the manual are met.

Can the owner unplug DC connectors to test the inverter?

No. PNNL states that modules are live in daylight regardless of disconnect status, and that only trained authorized people should connect or disconnect wiring. Use the visible AC and DC disconnects in the posted order, and leave connector work to that person.

Sources and references

Shutdown order, live modules, production checks, the roughly 10 percent insolation band, filters, and logs: Pacific Northwest National Laboratory. Islanding, ride-through priority, and UL 1741: NREL/Sandia IEEE 1547-2018 primer, 2022. Ride-through and detection tests: National Laboratory of the Rockies. Field temperatures: Thermal study of inverter components. Fans as a replacement item, and climate: 2013 thermal workshop notes and cooling-fan reliability assessment.