Solar inverter isolation fault on wet mornings: what the error means and when it stops being weather

What the inverter is testing before it lets the panels export

Both the SolarEdge and SMA documents are written for transformerless inverters, which have nothing inside that galvanically separates the panel side from the grid side. SolarEdge's note puts the consequence plainly: in a system with a transformer-less inverter, the DC is isolated from ground, and anything that lets DC current leak to protective earth is an isolation fault. The isolation it is talking about is the plastic and glass between live parts and anything grounded: the module backsheet and frame seal, the jacket on every PV cable, the body of every connector, the housing of each optimizer, and the inverter's own internals.

The manufacturers list the usual suspects in nearly the same words. SolarEdge names modules with defective module isolation, unshielded wires, defective power optimizers and an internal inverter fault. SMA's document says the cause is often defective insulation or a faulty plug connection on the modules or in the DC wiring, adds that rain or dew can make the event appear temporarily, and notes that strong wind can push a live cable with damaged insulation against a grounded part of the racking. That last point matters for owners: a fault that appears only on windy, wet days is still a physical defect somewhere on the roof.

The inverter does not measure each module separately. SolarEdge describes the number it compares with its limit as the total combined isolation resistance, and a combined resistance behaves like parallel paths: every module, cable and connector adds another route to ground, and the total is always lower than the weakest single path. That is the arithmetic behind the morning pattern. A healthy installation has a large margin; one wet, cracked connector can eat most of it; and the remaining paths decide whether the total lands above or below the line.

Put numbers on it. SolarEdge's troubleshooting procedure treats a homerun cable, or an optimizer and its module tested together, as healthy when an insulation tester at 500 V reads 200 MΩ or more. Twenty such paths in parallel combine to 10 MΩ (200 divided by 20), still about 17 times the 600 kΩ trip line. Now let one connector take on water and fall to 700 kΩ for an hour. Combined with the other 10 MΩ, the total drops to roughly 654 kΩ, inside the 10 percent band that SolarEdge calls borderline. A little more water and the inverter stays off; a little sun and it starts. These are illustrative values worked from the note's pass line, not measurements from any house, but they show why one defect, not a whole wet roof, is usually the story.

Four manufacturers, four names for the same leak

Each manufacturer names the event differently, publishes a different amount of detail, and handles the restart differently. The difference that matters most to an owner is the last column: some inverters retry on their own and some latch the fault until a person clears it.

Three practical readings come out of the table. First, a SolarEdge or SMA owner can see the fault clear by itself and wrongly conclude it is fixed; the documents describe self-clearing as normal for moisture-driven faults and still direct a qualified inspection. Second, an Enphase owner will not see that self-clearing at all, so a red unit after a wet night stays red on a sunny afternoon; the microinverter section explains why. Third, the published limits differ by a factor of almost two between SolarEdge's single-phase and three-phase products, so a resistance value read off one system means nothing on another. A house on an ordinary 240 V split-phase service uses a single-phase unit; the three-phase line matters for buildings served at three-phase voltages.

The Fronius row describes an older design philosophy. The Primo installation instructions say the inverter is designed for module grounding at the negative pole as standard, through a GFDI fuse, and that when the modules are grounded the inverter's insulation monitoring is deactivated. On a system set up that way, a leak shows up as a ground-fault event rather than a failed morning resistance test. The same document says that for ungrounded modules the GFDI fuse is removed and the grounding settings changed before operation, so the installer's commissioning record, not the brand name, tells you which kind of protection your system uses.

Wet mornings only, dry days too, or a falling number: sorting the pattern

The useful question is not whether the fault happened but what it does over two or three weeks. Moisture changes the resistance of a weak spot; it does not create one. So the weather pattern of the fault tells you how bad the defect is, and how fast to act.

SolarEdge gives the clearest numeric rule an owner or installer can apply. Its note says to check the isolation value in the monitoring platform (the inverter's info panel lists Last Isolation Value). If that value is borderline, which the note defines as within 10 percent of the limit, or below the limit, the fault should be troubleshot on site. If it sits at least 10 percent above the limit, the note says it is better to wait until the error recurs. On a single-phase inverter with a 600 kΩ limit, that 10 percent line is 660 kΩ. Your installer can read the value remotely in most cases; ask for it, and for its history if the platform keeps one.

The first row is the one most owners are in, and it is the one most likely to be ignored. A system that loses an hour or two after rain looks like it is working. The isolation value says otherwise, because a weak spot that fails when wet is the same weak spot that SMA warns can turn into a ground fault during operation. If the installer reports a value well above the 660 kΩ line and no recent events, waiting for a recurrence is what SolarEdge itself recommends; if the value is close to the line, waiting only moves the visit to a worse day.

Recording it from the app so the technician comes on the right morning

SMA's corrective steps begin with something an owner can do: check the frequency and duration of the event message in the inverter's interface or the monitoring portal. The same document then asks that the technician's measurements be taken under the same ambient conditions and at the same time of day as the error, ideally while the event is active. That instruction changes how you should report the problem. A technician who visits at 2 p.m. on a dry day will often find nothing, because the leak is not leaking. The record below is what gets the visit booked for a wet morning instead.

Keep the log for at least two weeks that include some rain or heavy dew. It costs nothing and it answers the questions any installer will ask in the first five minutes of the call.

1. Note the time of first production each morning from the daily power curve, and the time of sunrise. A normal system starts producing soon after first light; an isolation fault shows as a flat zero line that jumps up once the array dries.
2. Screenshot or export each event with its exact wording and code: 18x86 or 2x19 on SolarEdge, 3501 on SMA, DC Resistance Low - Power Off on Enphase. Write the code exactly as shown; a transposed digit sends the installer down a different path.
3. Write down the weather for the previous night and the morning: rain, dew, fog, frost melt, or dry. A phone weather app's history is enough.
4. Note anything that happened on the roof or around the array since the last clean month: roof repairs, gutter work, a storm, squirrels or birds nesting under the panels, a satellite dish installed nearby.
5. If your monitoring shows per-panel or per-string data, note whether the whole system starts late or only one string or unit. A single latched microinverter is a different call from a whole string inverter that waits for the sun.
6. Send the record to the installer in writing with the system's serial or site number, and ask two specific questions: what is the current isolation value, and can the visit be scheduled for a morning after rain.

This record also belongs in the monthly routine described in the solar maintenance guide, which treats reading the event log as the owner's main upkeep task. An owner who already checks the log monthly sees a late-start pattern long before it becomes an all-day outage.

Two things the record must not include are measurements you took yourself and anything that involved the DC disconnect. Switching a rooftop or inverter DC disconnect does not make the wiring safe to touch on a system with a ground fault; Fronius's instructions point out that on a grounded system the DC disconnect does not separate the modules from ground. If the inverter's display or app tells you to leave it off, leave it off.

Microinverters: the fault that stays red after the roof dries

Microinverter systems do not run one insulation test on a long string at sunrise. Each microinverter checks its own module. Enphase's IQ8 Series manual describes an insulation resistance sensor in every unit that measures the resistance from the positive and from the negative PV input to ground; if either falls below a threshold, that microinverter stops producing and raises a DC Resistance Low - Power Off condition. The manual lists the causes as defective module insulation, defective wiring or connectors, moisture ingress or a similar problem.

The behaviour after that is the opposite of the string-inverter pattern. The manual says that although the cause may be temporary, the condition persists until the sensor is manually reset, and that an IQ Gateway is required to clear it. The status LED on the unit stays solid red and the gateway keeps reporting the fault. A homeowner app may show the affected panel as producing nothing on a sunny afternoon, long after the dew that tripped it has gone.

Clearing it is an installer task. The manual describes two routes, both in Enphase's installer tools: a reset command for the affected serial number in the Enphase Installer Platform, or a clear message sent from the Enphase Installer App. It also says the condition will not clear if the cause is still present, and to contact the installer if it persists. So the owner's report is short: which panel, since when, and what the weather was the night before. If a unit clears and latches again after the next wet night, the cause is physical and needs a roof visit.

The cost of waiting is smaller per unit than on a string system, which is exactly why these faults sit unnoticed. The Department of Energy's Q1 2025 residential cost benchmark models its 8 kW system as twenty 400 W modules, each paired with a 350 W microinverter. On that system, one latched unit removes about one twentieth of the array's output, and nothing in the daily total looks dramatic. Per-panel monitoring is what catches it; the microinverter and string inverter comparison explains how that per-panel view is one of the real differences between the architectures.

Why a ground fault during production can destroy the inverter

The morning test exists to stop the inverter connecting a leaking array to the grid, and SMA's technical information explains why that matters so much for transformerless designs. With no galvanic isolation between the DC and AC sides, a ground fault that develops while the inverter is feeding in lets current driven from the grid side pass through the inverter's grid bridge, and the document says that current can exceed 1 kA. The inverter's residual current monitoring can detect the fault and open its grid relays, but SMA says that in a fault the disconnection cannot happen fast enough to prevent damage, because the current rises far faster than the components can tolerate.

Two consequences follow, both in SMA's words. The damage can show up as other event codes (64xx or 77xx) or as an inverter with all its LEDs off. And because the inverter and its components meet the standard requirements, damage caused by a ground fault during operation is not treated as an inverter fault, so it is not covered by the warranty. A weak spot that only fails the morning test is a warning that this risk is present. Fixing it while it is still a late start is a service visit; fixing it after it has become a fault during production can be a new inverter at your own expense.

There is a personal safety side as well. SolarEdge's note warns that a ground fault can put hazardous voltages or currents on normally grounded conductors or exposed metal, and Fronius's US instructions warn that a normally grounded conductor can become ungrounded and live during a ground fault, adding that an electric shock can be fatal. Rails, frames and conduit that are normally safe to brush against may not be while a fault is present. That is why the owner's role stays on the ground, reading the app, and why the rules on DIY solar work are worth reading before anyone suggests a self-repair.

What a thorough service visit looks like

Knowing the procedure lets you judge the visit. SolarEdge's note lays out the method in order. The technician turns the inverter's switch off and waits for the DC voltage to fall to a safe level, which the note gives as under 50 V on the display or five minutes for the capacitors to discharge. Strings are then reconnected one at a time to find the one that brings the error back. On a single-phase SolarEdge system, the display should show roughly 350 V DC once a healthy string is connected and the inverter is on.

Within the faulty string, SolarEdge inverters can point at the location. The Isolation Status diagnostic screen shows the resistance in kilohms and a percentage that marks where along the string the leak sits, from 0 percent at DC positive to 100 percent at DC negative. The note's own example: a string of 15 modules showing 55 percent puts the fault near module 8 (15 times 0.55 is 8.25), counted from the positive end. The location is accurate to about one module either way on single-phase inverters and two on three-phase, and if nothing is found there the note warns the string may have more than one fault.

The insulation tester is the next tool. With the cables disconnected, the technician tests the homeruns between the array and the inverter at 500 V; under 200 MΩ, the note says, the cable is faulty and is replaced. Optimizer and module pairs are tested the same way, and a pair under 200 MΩ is split to find whether the optimizer, the module or both are at fault. For the module itself, the note refers technicians to the IEC 61215 module insulation test and to the module maker's own specifications before testing.

SMA's version adds measurements a technician should take before replacing anything: open-circuit voltage of each string compared with the design, the voltage from DC positive to protective earth and from DC negative to protective earth (a stable reading close to the full string voltage indicates a ground fault), and an insulation resistance test on each string with the string short-circuited and without exceeding the modules' maximum system voltage. It finishes with a visual inspection of every module, cable, plug and module-level device for porosity, cracks, crushing, deformation or discoloration, and suggests an infrared camera to help.

What the service report should name

• The isolation or insulation value before the work, and after it, with the inverter's limit beside it.
• Which string was faulty, found by connecting strings one at a time.
• The location within the string, from the inverter's isolation status screen where it has one.
• Insulation tester readings for homeruns and for each component tested, with the test voltage used.
• The part replaced or repaired: connector, cable, optimizer, microinverter or module, with its serial number.
• Whether the visit took place in conditions like those when the fault appears, or why not.
• Who the repair was billed to: product warranty, workmanship warranty, lease owner or you.

The price of a lost morning, from a Denver production model

A moisture-driven isolation fault costs energy before it costs hardware, and the energy it costs is the early part of the day. To size that, this site's saved PVWatts run for an 8 kW DC system in Denver (south-facing, 20 degree tilt, DC to AC ratio 1.2, 14 percent losses, PVWatts version 8.5.0, typical-year weather, saved October 30, 2026) was split by hour.

The table shows the average production in the model's hourly output before 10:00 and its share of the day. These are averages over every day of the typical year, sunny and cloudy alike. A wet, overcast morning produces less than the average, so the figures are an upper marker for what a late start costs, not a forecast.

Summer mornings carry more of the day because the sun rises early and high; winter mornings carry less. The modelled year totals 12,760 kWh of AC output. That annual figure, and the monthly averages above, are what the illustration below works from.

Thirty late mornings against three dead weeks

Two owners have the same 8 kW system, modeled on the Denver run above. These are labelled illustrations with every input stated; no real system is described.

The first owner has a moisture-driven fault that holds the inverter off until 10:00 on ten mornings each in March, April and October. Using the monthly averages, the upper marker for the lost energy is 10 times 8.45 kWh, plus 10 times 10.77 kWh, plus 10 times 7.88 kWh: 84.5 plus 107.7 plus 78.8, about 271 kWh. Against the model's 12,760 kWh year, that is about 2.1 percent, and the real figure is lower because wet mornings are dimmer than average ones.

The second owner ignored the same pattern until the defect failed in dry weather in June, and the system sat off for three weeks while a visit was booked and parts arrived. At the June average of 42.6 kWh a day, 21 days is about 895 kWh, or 7.0 percent of the modelled year, more than three times the first owner's loss in less than a month. That is before any hardware cost, and before the SMA warning that a fault during production can destroy the inverter outside warranty.

Turning energy into dollars depends on your rate and export rule, which the net metering guide explains. As a scale marker only: every 10 cents of value per kWh makes the first owner's loss about $27 and the second owner's about $90. For comparison, the Department of Energy's Q1 2025 benchmark carries operation and maintenance at $34 per kW DC per year, about $272 a year for 8 kW, against a modeled system price of $2.95 per watt DC, or $23,600 (2024 dollars). The lost energy is small next to the system; the hardware it puts at risk is not.

When the fault has a date: roof work, storms, expansions and older grounded systems

A fault that starts on a known day is easier to find and often easier to assign. Roof repairs and panel removal are the classic cause: cables get pinched under rails, connectors are left resting in gutters, and jackets are scuffed when modules are stacked. If the fault began after the array was taken down and put back, the contractor who did the reinstall is the first call, and the guide to removing and reinstalling panels covers what that work should include. If a roofer worked around the panels without removing them, the questions in the solar and roof warranty guide about who is responsible for penetrations and wiring apply here too.

Storms and wind fit SMA's description of mechanical influences pushing live cables with defective insulation against grounded parts. A fault that starts after a storm, with photos from the ground showing a hanging cable or a displaced module, may be a property damage question as much as a service one; the solar insurance guide covers how panels are usually treated under a homeowners policy. Animals belong in the same group: chewed cable insulation under the array is a physical defect, and the fix starts with the cable, not the inverter.

Expansions change the arithmetic. Because the inverter compares a total combined resistance with its limit, adding modules adds parallel paths to ground and lowers the total even when every new part is healthy. A system that passed comfortably can sit closer to the line after an expansion, and a marginal connector that was harmless before can start failing on wet mornings. The expansion guide covers the checks an addition triggers; ask for an isolation value before and after the work.

Older systems grounded at the negative pole, like the Fronius Primo configuration described earlier, signal a leak through their ground-fault protection rather than a morning resistance test. The GFDI fuse in those inverters is part of the protection and replacing it is not an owner task; the Fronius instructions say a ground fault must be repaired before operation is resumed. If you own an older system and the installer has closed, the guide to a solar company going out of business explains how to find a new service provider and what records to bring.

Warranty, workmanship and the lease contract

Isolation faults sit between three kinds of cover, and the cause decides which one applies. A failed optimizer or microinverter is a product warranty claim against the manufacturer of that part, often filed by the installer. A cracked module backsheet is a module warranty question. A pinched cable, an unmated connector or a cable left in standing water is workmanship, and the installer's workmanship warranty, if the contract has one, is the cover to read; its length and exclusions are in your contract, not in any manufacturer document.

SMA's exclusion is the one to keep in mind when deciding how quickly to act: damage from a ground fault during operation is not covered by its inverter warranty. That makes the speed of the first report part of the owner's protection. A dated, written report of a late-start pattern, with the codes attached, shows the defect was raised while it was still a warning.

If the system is leased or under a power purchase agreement, the company that owns the equipment is responsible for repairs, and the contract's performance terms may cover the lost production. The lease, purchase and PPA comparison explains who holds which obligation. When an inverter is beyond repair, the inverter replacement guide works through the repair, like-for-like and architecture-change options and builds a cost envelope from the federal benchmark.