Commercial Solar Underperforming? 9 Causes and How to Check Them
Updated 23 October 2026 · SEO Dons Editorial
What should actual output be compared with?
Start with a like-for-like expectation. IEC TS 61724-3 defines a procedure for comparing the measured energy of a specific PV system with the energy expected for the same system under actual weather, and counts the times a system is not operating as an availability metric. PVGIS gives a long-term average for a place and layout, not weather-matched, and it leaves out snow, dust and dirt, and partial shadowing, so a gap can be real without being a fault. The losses guide shows what an expectation already allows for.
Weather alone moves single months a lot. For a south-facing 10 degree array in Birmingham, PVGIS's standard deviation for the monthly output due to year-to-year variation is between 7% and 17% of the monthly average, against about 3% for the annual total.
| Month | Average output (kWh per kWp) | Year-to-year standard deviation (kWh per kWp) | Deviation as share of average |
|---|---|---|---|
| January | 23.4 | 3.1 | 13% |
| February | 37.6 | 6.4 | 17% |
| March | 71.1 | 10.7 | 15% |
| April | 101.7 | 14.1 | 14% |
| May | 118.6 | 13.2 | 11% |
| June | 121.8 | 13.2 | 11% |
| July | 121.4 | 13.0 | 11% |
| August | 103.8 | 9.1 | 9% |
| September | 79.7 | 5.8 | 7% |
| October | 49.2 | 5.4 | 11% |
| November | 27.8 | 4.1 | 15% |
| December | 19.3 | 3.0 | 16% |
| Year | 875.3 | 28.5 | 3.3% |
Method: PVGIS 5.3 PVcalc, PVGIS-SARAH3 radiation, 1 kWp, south-facing, 10 degree tilt, system losses 14%, run in October 2026. Shares are illustrative arithmetic.
As a hypothetical example, a 200 kWp array that reported 150,000 kWh over twelve months against 175,000 kWh expected from the table (200 × 875) is 14% short, more than four times the 3.3% annual spread, so a cause is likely. The same shortfall in one winter month would prove little. So judge a system over twelve months, or by performance ratio, which Sandia notes reads lower in warm months. A sustained fall against the same month last year, adjusted for sunshine, or one string falling behind its neighbours, is the signal to investigate. The monitoring guide covers the data to collect.
What are the nine causes, and who fixes each?
| Cause | Typical pattern in the data | Usually fixed by |
|---|---|---|
| 1. Inverter trip or fault | One inverter low or at zero, with an event code | O&M contractor or inverter maker |
| 2. String or fuse fault | One string below its neighbours | O&M contractor |
| 3. Export limitation curtailment | Output capped on sunny, low-demand hours | Designer and installer, within the network agreement |
| 4. Voltage-rise disconnection | Trips or reductions with grid overvoltage events | Network operator and designer |
| 5. Soiling | Slow decline, recovery after rain or a clean | Cleaning contractor |
| 6. New shading | Dip at the same time each day in the same strings | Owner and designer |
| 7. Module defect | Persistently low strings, hot spots on thermography | O&M contractor, module maker |
| 8. Degradation | Slow yearly decline across everything | No repair; a warranty claim only if faster than warranted |
| 9. Meter or CT error | Impossible values, such as export at night | Installer or metering contractor |
The patterns are typical, not diagnostic: several causes look alike in a chart, which is why the event log and a site check matter.
What do manufacturers and standards say about each cause?
- Inverter faults. Manufacturers publish event codes. SMA's manual for its Sunny Tripower CORE2 lists, for example, a temperature event (inverter temperature excessively high and outside the safe range; check and clean the air inlets) and a low insulation resistance event, generally poor insulation of module or cable to ground, or rainfall and a moist environment. IEC 62446-2 covers corrective maintenance and troubleshooting.
- String and fuse faults. Sandia gives a blown fuse, broken wire or failed connector causing an open circuit as the simplest DC failure. SMA's string event tells the operator to check the connection state of the string. The IEC 62446-1 commissioning tests are the baseline for later string measurements.
- Export limitation. NGED says a scheme measures power at the exit point and either restricts generation or storage output or raises demand to keep within the agreed export capacity, and that it must fail safe. Settings left on can also cap output: Solis's low-power checklist includes switching off the export controller's fail-safe mode where none is installed. The G100 guide explains the schemes.
- Voltage rise. Regulation 27 of the Electricity Safety, Quality and Continuity Regulations 2002 declares a low-voltage supply at 230 volts with a permitted variation of 10% above, which is 253 volts. SMA's grid overvoltage event says the grid voltage is higher than the set value and advises measuring it, contacting the grid operator if it exceeds that value, and checking the AC cable cross section.
- Soiling. PVGIS excludes dust and dirt. The cleaning guide covers how to confirm it and when cleaning pays.
- New shading. PVGIS has no way of modelling partial shadowing, and Solis's checklist lists shadows on the panels among the first checks. The shading guide shows the cost.
- Module defects. Sandia says that in a module with three bypass diodes, open-circuit failure of a cell or short-circuit failure of a bypass diode reduces module output by one third. IEC TS 62446-3 defines outdoor infrared thermography of operating plants, and IEC TS 62804-1 covers testing modules for potential-induced degradation, an effect of high-voltage stress.
- Degradation. PVGIS cites typical losses of about 0.5% of power a year. The lifespan and degradation guide sets that against warranty lines.
- Meter or CT errors. Solis says a reversed current transformer misreads import as export and the reverse. With an export limit of 100 kW, importing 100 kW is misread as exporting it, and the controller ramps output further, causing trips. Mismatched phases make channels read about 50% low, or negative.
What should be checked first, and who should do it?
- Confirm the data: compare the inverter display, portal and generation meter.
- Compare like with like: same period, sunshine-adjusted, and the same month last year.
- Find the shape: one inverter, one string, a time of day, or a season.
- Read the inverter event log against the maker's documentation.
- Match the cause to the right party, and keep the findings for any warranty claim; the warranty guide shows what claims need.
Roof and DC-circuit fault-finding is work for competent people. HSE says all roofs, once fixed, should be treated as fragile until a competent person has confirmed otherwise, and regulation 16 of the Electricity at Work Regulations 1989 bars anyone from work needing technical knowledge to prevent danger unless they have it or are suitably supervised. The commercial PV systems page sets out what a specification and handover should include, and you can send an enquiry through the site's form to request quotes that cover monitoring and fault response.
Sources
- International Electrotechnical Commission: IEC TS 61724-3:2016; IEC 62446-1:2016; IEC 62446-2:2020; IEC TS 62446-3:2017; IEC TS 62804-1:2025 (publication pages), checked 22 October 2026.
- European Commission Joint Research Centre (PVGIS): Data sources and calculation methods; PVcalc run, south 10 degrees, Birmingham (PVGIS 5.3, 1 kWp, system losses 14%), checked 22 October 2026.
- Sandia National Laboratories, PV Performance Modeling Collaborative: DC component health; Annual yield; Performance ratio, checked 22 October 2026.
- SMA Solar Technology AG: Sunny Tripower CORE2 STP 110-60 operating manual, grid overvoltage event; temperature event; insulation resistance event; string event (manufacturer pages, own product only), checked 22 October 2026.
- Solis: Smart meter phase rotation and CT polarity verification; Low power generation (manufacturer pages, own product only), checked 22 October 2026.
- National Grid Electricity Distribution: Export limitation, checked 22 October 2026.
- legislation.gov.uk: Electricity Safety, Quality and Continuity Regulations 2002, regulation 27; Electricity at Work Regulations 1989, regulation 16, checked 22 October 2026.
- Health and Safety Executive: Fragile surfaces, checked 22 October 2026.
Common questions
How far below the PVGIS figure is normal?
For a single month, PVGIS's own standard deviation for the Birmingham example runs from 7% to 17% of the monthly average, and for a full year it is about 3%. A gap beyond that, repeated across months or against the same month last year after allowing for sunshine, deserves investigation. PVGIS also omits snow, dust and partial shading.
Should the owner go on the roof to look?
No. HSE says all roofs should be treated as fragile until a competent person has confirmed otherwise, and DC circuits are live in daylight. Start with monitoring data and the inverter event log, and leave roof and electrical checks to competent people.
Can an export-limited system underperform on purpose?
Yes. A limitation scheme restricts generation or storage output to stay within the agreed export capacity, so some sunny hours are capped by design. The yield model should state the expected curtailment, and the guide to G100 schemes explains how they work.
When does low output become a warranty issue?
When output falls faster than the module's performance warranty line, or an inverter fails inside its warranty term. Soiling, shading, curtailment and metering errors are operating or design matters instead. Keep the data that shows the pattern, because claims depend on evidence.