Capacity factor vs performance ratio: how to read a solar park's numbers

Capacity factor divides a solar park's energy by what its rated capacity could produce all year. Performance ratio divides it by what the sunlight on the panels could produce. For a 1 kilowatt-peak (kWp) fixed-tilt system at Szolnok, the European Commission's Photovoltaic Geographical Information System (PVGIS) gives 14.3% on the direct current (DC) basis and 78.75% respectively.

Both figures are model results, not measurements. On the inverter rating, which is alternating current (AC), the capacity factor would be higher.

What does capacity factor measure?

Capacity factor (CF) is the energy a plant delivers divided by the energy it would deliver at rated capacity in every hour of the year. The US Annual Technology Baseline (ATB) defines it the same way.

CF = E / (P_rated × hours)

E is energy in MWh, P_rated is capacity in MW and a non-leap year has 8,760 hours. Photovoltaic (PV) texts also call it the capacity utilisation factor (CUF). The International Energy Agency's Photovoltaic Power Systems Programme (IEA PVPS) Task 13 report T13-28 says CUF is usually 10% to 20%, higher with trackers, strong sun and high DC/AC ratios. It depends heavily on weather.

Why does the AC or DC basis change the capacity factor?

The energy stays the same, but the capacity in the denominator changes. Modules are rated in megawatts-peak (MWp), their DC output at standard test conditions. Inverters are rated in MW of AC. Developers often install more DC than AC, and the ratio is called the DC/AC ratio or inverter loading ratio (ILR).

The ATB reports capacity factor in AC terms and calls the ILR "a design choice that influences the capacity factor". Lawrence Berkeley National Laboratory (LBNL) also uses AC capacity, to compare solar with other generators. Its 2025 edition finds the highest-ILR quartile of US plants averages about 1 percentage point more capacity factor than the lowest. Its 2019 edition notes that oversizing can lower the DC-basis figure, because clipping removes energy at peak output.

SolServices' Lumen Park Szolnok shows the scale: 138.35 MWp DC against 96 MWac, a DC/AC ratio of about 1.44 (138.35 / 96). The figures come from a 2024 annual report of the park's parent group, and the company's project page gives 138 MW.

What does the performance ratio measure?

The performance ratio (PR) is the share of the energy the sunlight could have produced that reached the meter. The International Electrotechnical Commission (IEC) standard IEC 61724-1 covers PV performance monitoring, and PR is final yield divided by reference yield.

Y_f = E_out / P_STC            (final yield, kWh per kWp)
Y_r = H_POA / G_STC            (reference yield, hours)
PR  = Y_f / Y_r

E_out is AC energy at the plant output. P_STC is the DC module rating at standard test conditions (STC). H_POA is plane-of-array (POA) irradiation in kWh/m², and G_STC is 1 kW/m². These are the forms used in IEA PVPS report T13-28.

The reference yield is the number of hours at full STC power that the sunlight would have supported. PR therefore cancels most of the difference between a sunny and a dull year.

The second edition of IEC 61724-1, published in July 2021, replaced the 2017 edition. It added bifacial monitoring, updated irradiance sensor requirements and removed Class C monitoring systems. A sensor maker's summary of the standard says Class A, for utility-scale plants, requires ISO 9060 Class A pyranometers with dew and frost mitigation.

What is the temperature-corrected performance ratio?

It is the PR with the seasonal effect of module temperature removed. Output falls as cells heat up, so raw PR dips in summer and rises in winter even when the plant is unchanged. The correction adjusts the expected energy of each interval to a reference temperature.

PR_Tcorr = Σ E_out,k / Σ [ P_STC × (G_POA,k / G_STC) × (1 + γ × (T_mod,k - T_ref)) × Δt_k ]

Here γ is the module's power temperature coefficient (negative, in %/°C, from the datasheet), T_mod,k is module temperature in interval k and Δt_k is the interval length in hours. IEA PVPS report T13-28 says T_ref can be 25 °C or the annual weighted module temperature, and that the correction must run at sub-daily resolution because daily or monthly averages give erroneous results.

The US National Renewable Energy Laboratory (NREL) published the annual-average variant in 2013 as the weather-corrected performance ratio. It takes the irradiance-weighted annual average cell temperature from the project weather file. The result is more consistent through the year, so it can serve as a performance guarantee. NREL's protocol still requires the availability guarantees to be met.

T13-28 adds that temperature correction is usually applied only in warmer countries such as Spain. The Szolnok example below swings 11.8 percentage points between January and July, so raw monthly figures need care in Hungary. The PV temperature loss calculator shows how a module's coefficient becomes heat loss.

What do the two metrics give for a Hungarian park?

A 1 kWp fixed-tilt system at Szolnok (47.18 N, 20.19 E) has a modelled capacity factor of 14.3% on its DC rating and a PR of 78.75%. Both come from two PVGIS outputs, annual energy and in-plane irradiation.

The PVGIS API is free and needs no key. The request used crystalline silicon on a free-standing fixed mount, the optimal slope and azimuth, and the 14% system loss PVGIS proposes as its default. The manual says it covers cables, inverters, and dirt or sometimes snow on the modules, and notes that ageing lowers lifetime-average output by a few percent.

https://re.jrc.ec.europa.eu/api/v5_3/PVcalc?lat=47.18&lon=20.19&peakpower=1&loss=14&optimalangles=1&outputformat=json

PVGIS chose a 38° slope and an azimuth of -2° (2° east of south), using its PVGIS-SARAH3 satellite radiation data for 2005 to 2023. It returned annual energy (E_y) of 1,253.72 kWh and annual in-plane irradiation (H(i)_y) of 1,592.02 kWh/m².

CF (DC basis) = 1,253.72 kWh / (1 kWp × 8,760 h)  = 14.31%
Y_f           = 1,253.72 kWh / 1 kWp              = 1,253.72 kWh/kWp
Y_r           = 1,592.02 kWh/m² / 1 kW/m²         = 1,592.02 h
PR            = 1,253.72 / 1,592.02               = 78.75%

As a check, PVGIS reports total losses of 21.25%, and 100% minus 21.25% is 78.75%. PVGIS multiplies its losses: angle of incidence (-2.78%), spectrum (+1.34%), temperature and low irradiance (-7.05%) and the 14% system loss. This modelled PR is therefore the product of its assumptions, not a measurement.

On an AC basis the same energy is divided by 1/1.44 of the capacity, so 14.31% × 1.44 gives about 20.6%. PVGIS does not model clipping. Capping its hourly output for the same system at 1/1.44 of nameplate removes 1.5% of annual energy, which lowers the figure to about 20.3%. Hourly data hides short peaks, and LBNL notes that hourly modelling misses sub-hourly clipping, so the true loss is likely larger.

Capacity factor follows the sun: 6.1% in December, 19.7% in July. PR runs the other way, from 86.4% in January to 74.6% in July, because heat cuts output. Monthly PR is energy divided by irradiation, and monthly CF is energy divided by the hours in the month.

What values are normal for each metric?

For PR, a 2014 IEA PVPS Task 13 analysis found typical values above 80%, with some systems at 90% and a spread of 70% to 90% among new installations. The Szolnok model sits just below that at 78.75%. The right benchmark for a given park is the PR in its own energy model.

For capacity factor, no single value is normal, because it depends on the basis, mounting, ILR and site. The example gives 14.3% on the DC rating for a fixed tilt in Hungary. Trackers raise it by about 5 percentage points in LBNL's US data; see the trackers vs fixed tilt explainer.

What is each metric good for?

Capacity factor links MW of capacity to MWh of energy, so grid planners and revenue models start from it. It is technology-independent, which lets planners compare solar with other generators (T13-28).

PR strips out most of the irradiation effect, so it suits acceptance tests, warranty claims and operation and maintenance (O&M) reporting. T13-28 calls it the most commonly used key performance indicator (KPI) in Europe and lists it as contractually binding for asset owners, engineering, procurement and construction (EPC) contractors, O&M providers and consultants. SolarPower Europe's O&M guidelines say some providers still give PR guarantees, but best practice is availability and response-time guarantees, with at least 98% availability over a year.

What can distort the numbers?

Curtailment, downtime, clipping, sensor errors, snow, soiling and mixed capacity bases all move the figures.

• Curtailment. LBNL's capacity factors use net generation, which excludes curtailed energy, so curtailment lowers capacity factor. T13-28 lists curtailment among the periods contracts may exclude from PR. State which treatment a figure uses.
• Availability. Downtime lowers both metrics unless excluded. SolarPower Europe separates technical availability (all downtime) from contractual availability (which excludes causes outside the O&M provider's control).
• Clipping. At a high DC/AC ratio it lifts AC capacity factor but cuts PR, because T13-28 notes that more irradiation then means more clipped energy.
• Irradiance sensor quality. PR is only as good as its denominator. T13-28 calls reliable in-plane irradiance the most important measurement and warns that sensors near plant edges mislead. A sensor reading 2% low would lift this example's PR from 78.75% to about 80.4%. Dividing by horizontal irradiation (1,351.86 kWh/m² in a second PVGIS run) would give 92.7%, which is not comparable.
• Snow and soiling. The flat 14% loss does not model snow cover month by month, so the winter PR of 85% to 86% in the table is a model value. In most of Europe frequent rain means owners use a fixed cleaning schedule or none (T13-28).
• Different DC/AC ratios. Compare capacity factors only on the same basis and at similar ILR. Compare PR only for the same variant: raw, STC-corrected or annual-corrected.
• Weather and ageing. In PVGIS's hourly series, yearly totals for the same system run from 1,117 kWh/kWp (2010) to 1,309 kWh/kWp (2021). The ATB assumes 0.7%/year degradation; the PV degradation calculator shows what that does to output.

What this means for owners, grid planners and investors

Asset owners and O&M teams can track temperature-corrected PR next to availability, against the energy model's expectation for the same weather, and quote the irradiance source with every figure.

Grid planners and revenue analysts can use capacity factor on the AC rating, with the DC/AC ratio and curtailed energy shown separately. For how capacity itself is counted nationally, see why published Hungarian solar capacity figures differ.

Investors and lenders can ask four questions of any figure: AC or DC basis, which PR variant, which irradiance data, and which periods are excluded. The methods page sets out the definitions and data sources used across this site.

Limits of this explanation

The worked example is a PVGIS model of a generic 1 kWp fixed-tilt system with a flat 14% loss, averaged over 2005 to 2023. It is not measured output from Lumen Park Szolnok or any other park, and the park's ratings only size the DC/AC ratio. Real parks use trackers, bifacial modules and site-specific losses, which move both metrics.

The formulas follow IEC 61724-1 as set out by IEA PVPS Task 13, and contracts often define their own variants. The typical PR range comes from a 2014 analysis. This explainer cites no measured Hungarian utility-scale PR.