How to spot a failing solar panel before your installer does

If you have followed the Reading your solar charts: what the numbers actually mean and Estimates versus reality articles, you now have two useful tools: you can read the standard charts your monitoring app shows you, and you know how to tell normal variance from a real problem. The natural next question is what to do when the data does, after all, suggest something is wrong. This article is about the diagnostic step between "the numbers look off" and "I have called the installer". It is, in my experience, the most useful skill a solar owner can develop, because it transforms a vague worry into a specific question with a specific answer.

Before we go anywhere, the first thing worth knowing is that real failures are rare. A study by the US National Renewable Energy Laboratory found that roughly 5 out of every 10,000 modern panels fail, a rate of 0.05%. On a typical Belgian residential array of 14 panels, your statistical expectation is that none of them will ever fail outright in the next 25 years. The exception is the small but real category of systems where something goes wrong fairly early, often because of a manufacturing batch defect, an installation mistake, or a specific environmental factor. Catching those cases early is what good monitoring is for. Everything else is reading the chart correctly and not panicking.

What can actually go wrong

Failure modes in residential PV split usefully into four categories, in roughly increasing order of how easy they are to spot.

Balance-of-system failures are the easiest, because they are loud and obvious. The inverter stops producing. The DC isolator trips. The grid connection drops. The monitoring system shows zero production for an entire day or longer. These are not subtle, and most modern inverters will email or push-notify you the moment they happen. The fix is usually a service call to the installer, and the system is back up within days.

String-level failures are next. A solar array is typically wired as one or more strings of panels in series, and a failure that takes out one string (a tripped breaker, a corroded MC4 connector, a chewed cable from a curious marten) will drop a chunk of your production while leaving the rest of the system running. For a system with one string this looks identical to an inverter failure. For a system with two or more strings, you see a step-function drop in peak power that corresponds to losing one string's contribution, typically 30 to 50% of total output for a residential array.

Module-level failures are where it gets interesting. A single panel can fail in several ways: a bypass diode in its junction box shorts out (losing a third of that panel's output, which is hard to see in the aggregate data), the encapsulant browns over years (gradual percentage loss that mimics normal degradation), water ingress corrodes the cell contacts (gradual or sudden, depending on the leak), or a hailstone cracks a cell (immediate but sometimes barely visible). The most common failure types in systems installed in the last decade are hot spots and internal circuitry discolouration, which are both module-level issues that develop slowly enough that the data signature can be missed for years.

Cell-level degradation is the slowest and most insidious. Modern crystalline silicon panels are designed to lose 0.4% to 0.7% of their output per year as a normal part of ageing. A panel that is losing 1.5% per year is degrading abnormally, but the difference between 0.5% and 1.5% is invisible in any single year; it only becomes apparent when you compare year three to year one to year zero. This is where multi-year data, and the discipline of looking at it, earn their keep.

Reading the data signatures

Each failure category leaves a distinctive fingerprint in the production data, if you know what to look for.

A step-function drop in peak power is the signature of something going offline. Look at your daily peak power chart over the past month. If you see a clear line of values around 4.5 kW for three weeks, then 3.0 kW for the past week, something happened at that moment. The diagnostic question is what changed: a new shading object, a tripped breaker, a panel failure. The data will not tell you which, but the timing tells you when to look.

A gradual asymmetric decline is the signature of a single panel or string degrading faster than its neighbours. On a system with panel-level monitoring (Enphase microinverters, SolarEdge optimisers, or modern Tigo systems), this shows up directly: one panel reading consistently lower than the others. On a system with only string-level data (most older string inverters, anything from SMA, Fronius non-hybrid, or similar), it shows up as a slow drift downward in overall production that is not matched by your local weather. The harder case to diagnose is the second one. The diagnostic trick is to compare the daily curves on a clear day this year to a clear day at the same date last year. A small but consistent shortfall, repeated month after month, is the signal.

Lower peak power on bright clear days but normal output on cloudy days points to either inverter clipping at a lower threshold than before (which suggests the inverter is throttling itself due to overheating or a hardware fault), or a single failing panel that pulls down the entire string at high irradiance but contributes proportionally less of the loss at low irradiance. Either of these is worth investigating, because they imply the system is unable to capture the best conditions.

Shoulder loss in the daily curve, where the early morning and late afternoon production drops disproportionately compared to midday, usually indicates a problem with edge panels in the array, often those most exposed to weather. A failing panel at the east end of a string costs you morning production. A failing panel at the west end costs you afternoon. The midday peak can look almost normal because the failed panel's loss is averaged across many healthy ones.

Consistent week-of-year underperformance when compared across years suggests a fixed, repeating cause: a tree that has grown into a particular sun angle, a chimney whose shadow falls on a specific panel in winter, a piece of debris that lodged itself permanently somewhere. The calendar heatmap view makes this kind of multi-year pattern easy to spot, because the same week stays consistently darker in successive years.

Architecture matters: optimiser systems

The architecture of your installation affects both how failures manifest and how easily they are detected. A traditional string inverter system has one inverter at the bottom of the array, with panels wired in series strings. The number of active electronic components is small, and the failure modes are correspondingly limited.

Optimiser systems work differently. Most commonly SolarEdge, but also Tigo and Huawei optimiser variants, place a small DC-DC converter on each panel. This lets the system optimise each panel independently, handle mixed orientations on a single string, and provide panel-level monitoring as a built-in feature. The advantages are real, particularly for roofs with shading or multiple orientations.

The trade-off is that there are now many more electronic components exposed to rooftop temperature swings, humidity cycles, and UV. Each optimiser is a small failure point. Real-world reliability varies, newer SolarEdge optimisers (P505 and later) are noticeably more reliable than earlier ones, and installer reports range from near-zero failures to meaningful batch problems, but the architecture itself has more places where things can go wrong than a single string inverter does. Speaking from personal experience: I have had two batches of roughly five optimisers each fail over ten years on a SolarEdge installation, all replaced at no part cost under SolarEdge's 25-year optimiser warranty (the labour to access them was a separate negotiation). This is not catastrophic, and the warranty terms genuinely make it financially neutral, but it is not zero either.

The saving grace is that panel-level monitoring catches optimiser failures almost immediately. The diagnostic question shifts from "is something wrong with my system?" to "which specific panel is reading low?", and the answer is in the data within hours. For owners of optimiser systems, monitoring is less optional than for string inverter owners. A failed optimiser on a 14-panel array drops total production by around 7%, easy to miss in aggregated data, trivial to spot with panel-level visibility.

Microinverter systems (most commonly Enphase) share the same architectural logic: per-panel electronics, more failure points, panel-level monitoring built in. Reported failure rates for modern microinverters tend to be lower than for early optimiser generations, but the same monitoring-is-essential principle applies.

The honest summary: optimiser and microinverter systems are more failure-prone in absolute device count but far more diagnosable. The warranty terms generally absorb the financial impact. The real cost of the extra failures is the inconvenience of arranging the swap, not money out of pocket, provided you actually notice the failures, which is what monitoring is for.

Visual signs from the ground

The data narrows the diagnostic down. Visual inspection completes it. Most owners can do a useful first-pass inspection without ever climbing on the roof.

Discolouration: cosmetic versus diagnostic. Two distinct phenomena cause panel discolouration, and they have very different implications.

Uniform dulling across the whole array is what most owners notice first. After three to five years, the panels no longer have the deep, glossy blue-black appearance they had on installation day. They look a bit duller, sometimes with a slight haze that does not wash off in the rain. This is almost always cosmetic. It is a combination of the anti-reflective coating ageing, atmospheric deposition bonding to the glass surface beyond what rainwater can remove, and microscopic pitting from windblown dust. The production loss attributable to this kind of uniform ageing is typically under 2%, small enough to be hard to distinguish from normal degradation. You see this on almost every Belgian roof more than five years old, and it is not a sign that anything is wrong.

Discolouration on one panel that does not match its neighbours is the diagnostic version. A panel whose cells appear yellow, brown, or noticeably darker than the surrounding panels is showing encapsulant browning, the EVA layer that bonds the cells to the glass has degraded, usually from UV exposure or moisture ingress. This is real degradation, and the production loss tracks the visible change. The same logic applies to visible iridescent or rainbow patterns on a single panel: stress in the encapsulant that often precedes more serious problems.

The diagnostic question to ask is whether the appearance change is uniform across the array or specific to one or two panels. Uniform is cosmetic. Specific is diagnostic. A pair of binoculars is enough for a roof-level inspection from the ground or from a neighbouring upper-floor window.

Snail trails are another visible sign. These appear as faint, dark, branched lines running across the cell surface, looking like a snail has crawled over the panel. They are silver ion migration through microcracks, which means the panel has cell cracks that have started to propagate. Snail trails themselves are mostly cosmetic, but they are an indicator that the cell underneath is damaged. Production loss correlates with how extensive the trails are.

Visible cracks, chips, or impact damage on the glass are obvious problems but rare. A panel with a cracked face is at risk of water ingress over time, which will then accelerate every other failure mode. This is one of the few cases where immediate action is warranted, even before the data confirms a production drop.

Hot spots are usually invisible to the naked eye but show up clearly in thermal imaging. A cell or sub-region that runs significantly hotter than its neighbours is dissipating energy as heat instead of converting it to electricity, and over time this damages the cell further. We will come back to thermal imaging in the next section.

Burn marks, melted plastic, or discolouration at the junction box of any panel is a serious sign and warrants immediate attention. This usually indicates a failed bypass diode that has overheated, and in the worst case it can be a fire risk. If you see this, isolate the system and call the installer the same day.

Tools, what helps, what to skip

The diagnostic toolkit for residential PV has matured over the past five years. Some of it is genuinely useful for homeowners; some of it is not worth the money.

Binoculars are the most underrated tool. A €100 pair of 8×42 binoculars lets you inspect every panel on your roof from the garden in twenty minutes, far better than a mobile phone camera could. You look for discolouration, snail trails, cracks, dirt patterns, and anything that does not match the appearance of neighbouring panels. This is the single best DIY tool.

Consumer thermal cameras have become genuinely affordable. The FLIR ONE Pro or HIKMICRO Pocket2 attach to a smartphone for around €400 to €700 and give thermal images good enough to spot panel-level hot spots from the ground or a neighbouring window. They are not professional-grade, the resolution is modest and the precision is limited, but for a yes/no question of "is one panel significantly hotter than its neighbours under load", they work well. The trick is timing: thermal imaging needs the panels to be producing power, ideally on a clear sunny day with the sun high. A panel that is shaded or producing nothing looks the same as a healthy panel in idle.

Drones with thermal cameras are the next step up. The DJI Mavic 3 Thermal and similar prosumer models put a 640×512 thermal sensor in the air for around €4,000 to €6,000, which is overkill for one homeowner but reasonable for a solar installer doing inspections regularly. If your installer offers a thermal drone inspection as part of an annual service, it is usually worth taking up. The drone catches issues a ground inspection cannot, because it can fly above the panels and look straight down at them rather than at a steep angle.

IV curve tracers are the professional diagnostic tool. They sweep voltage across a panel or string while measuring current, producing a characteristic curve that reveals bypass diode failures, PID, mismatch, and other electrical issues. The good ones cost €3,000 to €10,000 and require training to interpret. This is firmly the installer's tool, not the homeowner's, but knowing it exists is useful when you want a definitive diagnosis of a suspected fault.

Insulation resistance testing is an electrician's job, not a DIY task. It checks for current leakage from the DC side of the system to ground, which can indicate water ingress, damaged cables, or PID. If your installer or electrician is doing an inspection, ask whether they include an insulation resistance test; it catches issues that production data alone cannot.

What to skip: consumer "solar panel testers" sold online for €30 to €100 that claim to diagnose panels. They are voltage or current meters dressed up with marketing, and the meaningful diagnostic tests need controlled conditions and trained interpretation. The €30 device that promises to "test your panels" mostly tells you whether they are producing voltage at all, which the inverter already does.

The diagnostic path with multi-year data

The most powerful diagnostic tool, if you have it, is multi-year history. The procedure is simple but takes a few minutes of focused attention.

Start by pulling up the year-on-year comparison for the past three or four years. Look at the annual totals: are they trending downward faster than the normal 0.5% per year? If yes, the system is degrading abnormally, and the next step is to find when the abnormal decline started.

Drill into the year when the abnormal decline began. Compare its monthly production to the previous year. The month where the divergence becomes visible is the diagnostic anchor point: whatever changed in the system happened during or just before that month. New shading? A storm event? A firmware update? A specific date that you remember as having had a problem? The annual report your monitoring tool generates (HelioPeak's annual PDF, PVOutput's statistics views, your inverter manufacturer's reports) is useful here because it has the comparisons pre-assembled.

Once you have a candidate date, check the daily peak power and daily curve shape from the days surrounding it. A failure that happened at a specific moment shows up cleanly: there is a "before" pattern and an "after" pattern, and they differ in a way that survives weather variation. A gradual degradation does not have such a clean transition, and points instead to a slow-developing issue (PID, encapsulant browning, cell-level degradation).

This is also where the Notes feature in HelioPeak we mentioned in Reading your solar charts: what the numbers actually mean becomes valuable. If you have been noting events, panel cleaning, inverter reboots, storms, scaffolding on the roof, the badges on your chart at the suspect date tell you instantly whether the timing matches a recorded event. If the failure date lines up with a stormy night, the diagnostic is mostly done; if there is no recorded event, you have ruled out the obvious causes.

Warranty enforcement: how the data becomes leverage

Modern solar panels come with two warranties that matter. The product warranty, typically 10 to 15 years for mainstream panels and 20 to 25 years for premium ones, covers defects in materials or workmanship. The performance warranty, typically 25 years (increasingly 30 on premium lines), guarantees a minimum power output over time, usually 80% to 87% of the original rating at year 25, with a defined annual degradation rate.

If your panels are dropping below the warranty curve, you have a claim. The challenge is proving it. Manufacturers do not accept "my system feels low" as evidence; they want production data over multiple years, dated photos of any visible issues, the original installation documentation, and a clear comparison to the warranty's degradation curve.

This is where consistent multi-year monitoring pays off. A homeowner who has been logging to PVOutput since installation can produce a year-by-year comparison that shows precisely when production dropped below the warranty floor. A homeowner who only has the manufacturer's app, with its limited history, may find the warranty claim much harder to substantiate.

In Belgium, you have an additional consumer-protection layer beyond the manufacturer warranty: the two-year legal guarantee under the Belgian Civil Code applies to all consumer goods, including solar panels. This is supplementary, not replacing the manufacturer warranty, but it gives you a stronger position during the first two years if the installer pushes back on a defect claim. After year two, you fall back on the manufacturer warranty, which the installer typically administers on your behalf.

The standard process for a warranty claim is to contact your installer first. They will normally do a remote diagnostic via their monitoring system, possibly come for a site inspection, and if they confirm a defect they file the claim with the panel manufacturer. The replacement panel is supplied at no cost. The labour to swap it is sometimes covered, sometimes not, depending on the installer's workmanship warranty and any premium support package. This is one reason why choosing an installer with a strong workmanship warranty matters as much as choosing premium panels.

A note on cleaning

Before assuming a production drop is a failure, it is worth ruling out the simplest possible cause: dirt. Soiling can cost 2% to 5% of annual production in Belgium, and more in dry climates or near agricultural land. The question of whether to actively clean is more nuanced than it appears.

The good news for most Belgian residential installations is that rain does most of the work. At the standard 30 to 45-degree tilt, water runs off panels efficiently, and the Belgian climate provides enough cleaning events through the year to keep most installations within acceptable soiling ranges. The first significant rain after a dry spell typically restores production by a percentage point or two, which makes professional cleaning a marginal economic decision for tilted residential installations.

The exceptions are worth knowing.

Flat or near-flat roof installations (below 15 degrees of tilt) do not self-clean effectively. Water pools at the bottom edge of each panel rather than running off, and a band of dirt accumulates along that edge over months. On such installations, soiling can cost 5% to 10% of annual production, and periodic cleaning genuinely pays back. Bird droppings are acidic, leave visible stains that do not wash off, and can cause hot spots underneath. Tree sap is sticky and similarly persistent. Heavy pollen seasons (April-May in Belgium), construction dust from nearby works, and agricultural or industrial particulate deposition can each push soiling losses into the worthwhile-to-address range.

When cleaning does make sense, the rules are simple. Never walk on the panels, the mechanical load causes microcracks that void warranties and accelerate degradation. Use a soft brush on an extension pole, distilled water or rainwater (tap water leaves mineral spots), and no abrasive or ammonia-based cleaners. Most professional cleaning services charge €5 to €15 per panel in Belgium, so a typical residential array runs €100 to €300 per cleaning. The maths usually favours cleaning every two to three years on flat-roof installations, rarely or never on standard tilted installations unless a specific contamination event has occurred.

The diagnostic relevance is straightforward: if your production drop is gradual and uniform across the array, soiling is worth checking before more serious diagnostics. A clean array after the next significant rain that returns to expected production tells you the issue was dirt. If production stays low after rain, you have ruled out one cause and can move on to the next.

When not to panic

Most "the system seems off" moments are not failures. The list of false positives I have seen most often:

• A cloudy fortnight in summer that drops a month's total by 20%. Normal weather variance.
• December production looking dreadful. Belgium produces 2% to 4% of the annual total in December. The system is fine.
• A single bad day after a stretch of good ones. Weather. Look at the week's total, not the day's.
• Soiling buildup after a dry spell. Will clear with the next significant rain. If you live near a road with lots of dust, or you have nesting birds, soiling becomes a recurring issue, but it is a maintenance question rather than a failure.
• A new neighbour's tree that has grown over the winter and now casts more shadow than it did a year ago. Real loss, but not a panel failure.
• A firmware update on the inverter that changed how data is reported. Compare units carefully; some updates change the timestamp grid, which can make sums look different without the underlying production changing.

The diagnostic instinct is to check weather first, check seasonal expectation second, check the local environment third, and only then start looking at the system itself.

A small closing thought

The point of all of this is not to turn solar owners into amateur diagnosticians. The point is to make the conversation with the installer better when something is actually wrong. A homeowner who says "my system is broken" gets generic reassurance and a slow service queue. A homeowner who says "my year-on-year specific yield has dropped 4% with no weather explanation, the divergence started in March 2025, the daily peak power dropped from 4.6 kW to 4.1 kW on clear days, and I can see what looks like discolouration on the third panel from the left" gets a fast, focused response and a likely warranty claim.

The data is the leverage. The visual inspection completes the picture. The installer brings the tools and the certification. Each of those three things on its own is incomplete; together, they catch the small number of real failures early enough that the loss is recoverable and the warranty is enforceable.

That is what monitoring is actually for. Not the daily glance, which is a pleasure. The diagnostic anchor, which is the genuine value.