Solar Plant Underperformance: A Diagnostic Sequence

01Establish the benchmark before touching anything

A system never delivers its STC nameplate in the field. Realistic models carry system losses – typically in the range of fifteen to twenty-five per cent – covering thermal derating, cabling, inverter efficiency, soiling, mismatch and availability. Seasonal variation alone can put winter output tens of per cent below summer in many climates.

Use an independent yield reference for the location (PVGIS is the common one in Europe; equivalents exist elsewhere) or the original design model, and compare twelve months of generation against it. A plant delivering roughly 80 to 100 per cent of a well-constructed reference is performing normally. Below about 75-80 per cent, the investigation is worth the labour. Above it, the problem is usually the expectation.

02Soiling and new shading

Contamination is the most common genuine cause and the cheapest to fix. Dust, pollen, bird droppings and industrial film typically cost somewhere between five and twenty-five per cent depending on the environment, and considerably more in arid regions or near construction and industrial sources. Note that oily industrial film behaves differently from loose dust and needs a different cleaning method.

Shading is the other half of this step, and it changes over a plant’s life: trees grow, adjacent buildings go up, and equipment gets added to roofs. The signature is output collapsing at the same clock time each day. Worth noting that a small shadow on one module can cost far more than its area suggests, because in a string the affected module limits the current of the whole string.

03Compare strings before you compare to the model

String-level monitoring turns a vague complaint into a location. On a clear day, strings on the same MPPT input should show similar current. If one reads near zero, the cause is usually discrete: a blown fuse, an open circuit, a disconnected or reversed conductor. If one reads consistently low but not zero, suspect mismatch, shading or a module fault.

Field experience puts module-side issues behind the majority of generation loss, with inverter and electrical faults a distant second. That ratio is a good argument for spending the first hour on the array rather than on the inverter.

04Understand which inverter behaviour you are seeing

Three inverter behaviours look similar on a dashboard and mean completely different things:

• Clipping – the array can produce more DC power than the inverter can convert, so output flattens at the inverter’s AC rating on the best days. This is usually a deliberate design choice and already priced into the model. It is only a fault if the model assumed it was not happening.
• Thermal derating – output falls during hot afternoons because the unit is protecting itself. Usually a siting and ventilation problem, not a defect. Check the derating curve against actual mounting conditions.
• Export limiting – the plant is being curtailed because the network will not accept more export at that moment. The plant is healthy; the constraint is external, and it should show as a plateau on sunny low-load days.

05Thermal loss and voltage rise

Heat is not a fault, but it is worth quantifying properly: modules lose roughly a third of a percent of output per degree above the rating point, and the exact figure depends on the cell technology – one reason we treat the temperature coefficient as a procurement specification in the module technology guide. Peak summer output landing ten to fifteen per cent below nameplate is normal and should be in the model.

Voltage rise is the other thermal-adjacent trap: long or undersized AC runs, or many inverters concentrated at one connection point, push the voltage toward the upper limit and the inverter reduces power in response. If the plant dips exactly when it should peak, check the cable sizing and the connection arrangement.

06Mismatch inside the array

Strings sharing an MPPT input should share orientation and tilt. Mixing roof faces, or mixing module types or ages on one input, means the stronger strings are pulled down toward the weakest. This is a design-stage error, but it is discovered in operations, and it is worth checking whenever a plant has been extended or partially repaired.

07When it really is a component defect

If the above checks are clean and the shortfall persists, move to component-level testing. Infrared scanning finds hot spots and failed bypass diodes quickly; I-V curve tracing quantifies how far a string has drifted from its expected curve; electroluminescence imaging reveals cell cracking and, in older plants, potential-induced degradation. These are the tests that convert a suspicion into a warranty claim, so capture the data in a form the manufacturer will accept.

08Symptom-to-cause reference

09Preventing it upstream

Most of the causes above are designed in rather than developed later – and several are caught by the pre-energisation tests in our commissioning checklist, particularly string polarity and voltage verification. The monitoring specification matters too: without string-level data, none of the diagnosis above can be done from a desk. See the monitoring requirements in the inverter and grid connection guide, and the equipment covered on the inverters and controllers page. For diagnostic support on a live plant, use the contact page.