Solar PV System Losses: Where the kWh Go
Updated 23 October 2026 · SEO Dons Editorial
How do losses build up between the sun and the meter?
A PV array starts with the sunlight that lands on the plane of its modules and loses energy in stages. PVsyst, a design package whose public documentation is a convenient map of a loss budget, lists the stages in simulation order: reflection at the glass, soiling, low-light efficiency, temperature, light-induced degradation, module quality, mismatch, ageing, optimisers, DC wiring, AC wiring, an external transformer, auxiliary consumption and system unavailability. Optimisers and an external transformer appear only where the design includes them.
Two rules stop the chain being misread. First, losses multiply. PVsyst says each loss is a percentage of the previous energy quantity, so grouped percentages are not the sum of the detailed ones, and PVGIS's worked example combines 3.7% reflection, 7.2% temperature and low-irradiance loss and a 14% system loss into 23.1%, not the 24.9% a straight sum would give. Second, every loss is measured against an ideal. PVsyst defines that ideal as 1 kWh of energy for each kWh of irradiance per installed kWp, so the first line of any budget is the irradiation reaching the module plane. The yield by region guide shows how that first line changes with place and layout.
What do the losses look like for a real array in Birmingham?
PVGIS publishes the first three steps as percentages. The table converts them into kWh per kWp for two layouts, a south-facing 35 degree pitch and a south-facing 10 degree flat-roof array, using Birmingham city-centre coordinates and a 14% system loss. The kWh steps are illustrative arithmetic on PVGIS's percentages, rounded, so they will not always sum to the last unit.
| Step | South, 35 degrees | South, 10 degrees |
|---|---|---|
| Irradiation on the module plane (kWh/m², equal to kWh per kWp at standard test conditions) | 1,193 | 1,093 |
| Shallow-angle reflection | -38 (-3.21%) | -45 (-4.12%) |
| Spectrum change | +21 (+1.81%) | +18 (+1.69%) |
| Temperature and low irradiance | -50 (-4.26%) | -48 (-4.47%) |
| System loss input (14%) | -158 | -143 |
| Annual output (kWh per kWp) | 968 | 875 |
| PVGIS combined loss figure | -18.87% | -19.90% |
Method: PVGIS 5.3 PVcalc, PVGIS-SARAH3 radiation data, 1 kWp, fixed mounting, system losses 14%, run in October 2026.
Three points stand out. The 14% input is the largest single step, about 158 of the 225 kWh for every kWp installed that disappear in the pitched case and about 143 of the 217 in the flat case, yet it is a bundled allowance rather than a measurement. The flat array loses more to reflection, which PVGIS ties to light striking the glass at shallow angles. And PVGIS's methods page lists snow, dust and dirt, and partial shadowing as effects its calculations do not include, so any of them comes on top of every figure above.
What sits inside the 14% system loss?
PVGIS explains the allowance in its methods page. After the module-level calculations, power still has to be converted to AC by the inverter, there are losses in cables, and modules are not all exactly the same power. PVGIS says it cannot calculate these effects, so the user enters them, and it recommends 14% once ageing is included. It cites a large study in which modules typically lose about 0.5% of power per year, which over 20 years leaves 90% of the original power at the end and 95% on average.
Other tools itemise the same ground. The table sets out the defaults and ranges two of them publish: PVWatts, the US calculator now run by the National Laboratory of the Rockies (the former NREL, renamed in December 2025), and PVsyst's documentation. They are modelling conventions to compare a proposal against, not measurements of any site.
| Loss | PVWatts default | PVsyst documentation | Who can change it |
|---|---|---|---|
| Soiling | 2% | Under 1% in moderately rainy climates and residential areas; several percent common in industrial zones | Operator, through cleaning |
| Shading | 1%, for little or no shading | Calculated from a 3D scene | Designer |
| Mismatch | 2% | Constant loss from a module-dispersion tool | Designer and module choice |
| DC wiring and connectors | 2% wiring, 0.5% connections | 1.5% default at STC, about 60% of that as annual energy | Designer, through cable sizing |
| Light-induced degradation | 1.5% | 1% to 3% or more for p-type wafers; n-type not affected | Buyer, through module type |
| Availability | 3% | Entered as outage periods, statistical | Operator, through monitoring |
| External transformer | Not listed | 0.1% of nominal power default for iron loss | Designer |
Inverter conversion is a separate line. One manufacturer's operating manual for a three-phase commercial inverter, the SMA Sunny Tripower CORE2 STP 110-60, lists a maximum efficiency of 98.6% and a European weighted efficiency of 98.4%, so plain conversion loss can be small. The larger inverter losses can come from limits. Sandia's PV Performance Modeling Collaborative explains that clipping occurs when the array's DC power exceeds the inverter's maximum input, and that some inverters also self-limit when internal temperatures rise because cooling has failed. The inverter types guide covers sizing the DC:AC ratio.
Which losses can a buyer actually change?
- Set by place and physics: reflection, spectrum and temperature. PVsyst puts the temperature effect at about -0.2 to -0.4% per degree above the 25 degree test condition, and module datasheets differ, but the climate is fixed.
- Set at design: tilt and orientation, the layout around plant and parapets, string design, cable lengths, module technology and inverter loading. The shading guide shows how large an obstruction has to be before it costs real energy, and PVsyst's documentation says light-induced degradation affects only conventional p-type wafers, which is worth asking about when comparing the cell types in the cell technology guide.
- Set in operation: soiling, availability and how quickly a fault is caught. The cleaning guide, the monitoring guide and the underperformance checklist cover these.
The operating losses are the ones a buyer inherits for the whole life of the system, and they are the ones a proposal is least likely to itemise. Itemised losses belong in the written specification that the commercial PV systems page describes, rather than in a single headline yield.
How do you compare loss assumptions between proposals?
Ask each proposal for its loss diagram, not just the annual kWh. Then check five things:
- The tool and weather data, for example PVGIS-SARAH3 or a commercial database, so the first line can be reproduced.
- Whether losses are itemised or bundled into one system-loss allowance such as 14%.
- The soiling, shading and availability assumptions, stated separately, because PVGIS leaves out dust, snow and partial shading.
- The DC:AC ratio and the clipping loss it produces.
- The degradation assumption for year one and for the average over the period used in the payback.
The same headline can hide different stories. As an illustration, one proposal that assumes 14% in total and nothing for soiling, and another that assumes a 10% system loss plus 2% soiling and 3% availability, end up within half a percentage point of each other (0.90 × 0.98 × 0.97 gives a 14.4% combined loss). The sums look alike, but only the second tells the buyer where to look when output falls short. The PVsyst report guide explains how to find these lines in a report, and the quote comparison guide covers the rest of a proposal. You can send an enquiry through the site's form if you want quotes that itemise losses this way.
Sources
- European Commission Joint Research Centre (PVGIS): Using PVGIS: frequently asked questions; PVGIS data sources and calculation methods; PVcalc run, south 35 degrees, Birmingham; PVcalc run, south 10 degrees, Birmingham (PVGIS 5.3, SARAH3 radiation data, 1 kWp, system losses 14%), checked 22 October 2026.
- PVsyst SA, online documentation: Loss diagram; Array and system losses; Array losses, general considerations; Soiling loss; Array ohmic wiring loss; External transformer losses; LID loss; Unavailability loss; Shadings (for its own loss model only), checked 22 October 2026.
- PVWatts Calculator, National Laboratory of the Rockies: PVWatts system losses categories, checked 22 October 2026.
- US Department of Energy: Energy Department Renames NREL "National Lab of the Rockies" (1 December 2025), checked 22 October 2026.
- SMA Solar Technology AG: Sunny Tripower CORE2 STP 110-60 operating manual, Efficiency (manufacturer page, own product only), checked 22 October 2026.
- Sandia National Laboratories, PV Performance Modeling Collaborative: Inverter saturation or clipping, checked 22 October 2026.
Common questions
Is 14% a realistic system loss for a commercial array?
PVGIS recommends 14% as an allowance covering system losses and ageing, and it is the figure behind the yield tables on this site. It is a planning input rather than a measured value. An itemised model that states soiling, availability and inverter losses separately gives a better basis for checking a proposal.
Why does the flat-roof array lose more than the pitched one?
In the Birmingham runs PVGIS gives a combined loss of 19.90% for the 10 degree array and 18.87% for the 35 degree one. Most of the gap is shallow-angle reflection, 4.12% against 3.21%. The flat array also receives less irradiation to begin with, which is a design trade-off rather than a loss.
Do PVGIS results include snow, dirt and shade from plant on the roof?
No. PVGIS's methods page lists snow, dust and dirt, and partial shadowing as effects its calculations do not include. Its terrain horizon also cannot see nearby objects, so those need separate allowances or a 3D shading model.
Why can the percentages not simply be added together?
Each loss applies to the energy left after the one before it. PVGIS's own example of 3.7%, 7.2% and 14% gives 23.1% rather than 24.9%, and PVsyst makes the same point about its loss diagram.