How to Assess Solar Module Degradation and Performance Ratio in the Field: IEC 61724-1/-2/-3 (Monitoring, Short-Term Capacity Test, Energy Test), Weather-Corrected PR, Year-on-Year Degradation Rates (RdTools; NREL's 0.5%/Year Median), IEC 60891 I-V Corrections, the Chinese Standards GB/T 39857-2021 and NB/T 32034-2016, and IEA PVPS Findings on Non-Linear and Climate-Specific Degradation

Short answer. Three questions are usually bundled under 'how is the plant performing': is the plant producing what the weather allows right now (a capacity test), did it produce what it should have over a period (an energy test), and is it losing output year by year (a degradation rate). The IEC 61724 family separates them: Part 1 (2021) is the monitoring standard that defines terminology, equipment and classes of monitoring systems and adds bifacial, irradiance-sensor and soiling updates; IEC TS 61724-2 is a short capacity test over a few relatively sunny days comparing measured against expected power (the 2016 edition read here has a 2025 successor); IEC TS 61724-3 is an energy test against expected production under actual weather, with unavailable time carried in an availability metric. The performance ratio itself — generated electricity divided by what the DC nameplate would have produced from the same sunlight — depends on weather as much as on the plant, so winter and summer PR measurements of the same plant may differ substantially; NREL's weather-corrected PR exists for that reason. For degradation, the field record is NREL's 2012 review of nearly 2,000 rates (median 0.5%/year, mean 0.8%/year, 78% below 1%/year, excluding short-term light-induced loss), and the method the RdTools documentation prefers is year-on-year, which yields a distribution of rates whose central value is the estimate and whose width gives the uncertainty, and which measures degradation rate, not the broader performance loss rate. Module-level I-V measurements need the temperature and irradiance corrections of IEC 60891, and China has its own efficiency and field-test standards, GB/T 39857-2021 and NB/T 32034-2016, whose texts this guide did not read. IEA PVPS Task 13 supplies the cautions: assumed rates are rarely system-specific and real degradation is non-linear, defect losses range from below detection to catastrophic, cut-cell cracks grow in the field, and each climate brings its own stressors.

1. Which document answers which question

CNWattHub reading: a capacity test and an energy test can disagree — a plant may pass a sunny-day capacity test and still miss its annual energy because of outages, soiling or curtailment that the availability metric and the energy test are designed to catch. Ask which test a 'PR guarantee' in a contract actually references.

2. Performance ratio: what it is and why it moves

NREL's 2013 report states the definition and the trap in one place: PR is the ratio of electricity generated to what would have been generated had the plant converted sunlight at the DC nameplate level; annual yield is plane-of-array insolation × nameplate × PR; and because PR is a function of both plant efficiency and weather, winter and summer measurements of the same plant can give substantially different values. IEC 61724-1 is the monitoring standard behind such measurements, and its 2021 edition updated the irradiance-sensor requirements.

CNWattHub reading: three consequences for an assessor. First, compare PR to PR only for like periods or after weather correction. Second, check the irradiance sensor before blaming the modules — the RdTools documentation names drift of irradiance sensors as a bias on degradation results and recommends re-running with satellite weather as a check. Third, treat a PR quoted without its period, sensor class and temperature treatment as a number without a method.

3. Degradation rate: the statistics and the method

The field record. NREL's 2012 review assembled nearly 2,000 degradation rates from modules and systems: median 0.5%/year, mean 0.8%/year (the distribution is skewed to high values), 78% of all data below 1%/year; the rates are long-term and exclude short-term light-induced degradation.

The method. The RdTools documentation describes a normalised-yield workflow — import, normalise with a performance metric, filter, aggregate (usually daily), filter again, analyse — and prefers the year-on-year approach: it yields a distribution of rates whose central tendency is the most representative estimate, with the distribution width giving the uncertainty by bootstrap. The documentation also draws a line an assessor should keep: the workflow estimates a degradation rate, not a performance loss rate, because PLR includes losses that the workflow deliberately filters out.

The cautions from IEA PVPS Task 13. Assumed lifetimes and degradation rates are in most cases not system-specific but averages from older systems or data sheets, and yield models typically assume linear degradation although field degradation is typically non-linear. The risk-quantification report grades a defect's expected loss, at component level and without mitigation, from none, through below detection limit (<2–3%), within warranty (<0.7–1%/year) and outside warranty (>0.7–1%/year), to catastrophic (>3%/year). The 2025 degradation report describes one in-field loss mechanism: laser-cut half and third cells carry micro-cracks at the cut edge that frequently grow into dendritic cracks in transport, installation and field loading; a cited figure is 0.2% power loss per dendritic-cracked half-cell, so about 25 such cells cost 5%.

CNWattHub reading: the arithmetic of a degradation claim matters as much as the number. A year-on-year median of 0.6%/year with a wide bootstrap interval from two years of data is not evidence of a warranty breach; a loss concentrated in cracked cells is a defect question, not a degradation-rate question. The warranty curve (linear, from the data sheet) and the field curve (non-linear, from the plant) are different objects, and Task 13 says so.

4. Climate and site

IEA PVPS Task 13's 2025 fact sheet (summarising report T13-32:2025) lists stressors by climate: in cold and snowy sites, snow loads that deform frames or break glass, embrittled encapsulants and solder joints, thermal cycling that causes microcracks, and snow reflecting 80–90% in the UV range at altitude; in hot and dry sites, soiling, high temperatures, thermal cycling, salt mist, UV and sandstorms; in hot and humid sites, moisture ingress, corrosion of junction boxes and connections, delamination, stronger dust adhesion and biological growth; and climate-specific design and O&M as the mitigation.

CNWattHub reading: a degradation benchmark from one climate is not a benchmark for another. When a lender's model carries the data-sheet linear rate for a desert or a tropical-coast plant, the assessor should ask for a site-specific rate or at least a climate-matched comparison set.

5. Module-level testing

Field I-V measurements are taken at whatever irradiance and temperature the day provides; IEC 60891:2021 defines the procedures for correcting them to other conditions and for determining the correction factors, with the measurement requirements themselves in IEC 60904-1 and its subparts. In China the on-site module test procedure is NB/T 32034-2016, an NEA industry standard drafted mainly by China Electric Power Research Institute, TBEA and State Grid Qinghai's research institute and centralised by the China Electricity Council, and the efficiency specification is GB/T 39857-2021, current since 1 October 2021. Neither text was read for this guide; a Chinese test report that cites them should be checked against the standard it names.

6. An assessment sequence built from the sources

CNWattHub's reading of the sources as a checklist:

1. Monitoring quality first. Sensor class, irradiance-sensor requirements, soiling measurement, bifacial handling per IEC 61724-1:2021; check sensor drift against satellite data.
2. Pick the test for the question. Capacity (61724-2, a few sunny days) versus energy (61724-3, period with availability); note the 2025 edition of Part 2.
3. Normalise the PR. Same period or weather-corrected; state period, sensors and temperature treatment.
4. Estimate degradation properly. Year-on-year with a bootstrap interval; report whether it is a degradation rate or a performance loss rate.
5. Benchmark against the field record. 0.5%/year median, 0.8%/year mean, 78% below 1%/year — long-term, excluding LID.
6. Separate defects from drift. Loss classes from below detection to catastrophic; cut-cell cracks as a defect mechanism.
7. Correct the climate. Match the comparison set and the model's rate to the site's stressors.
8. Module tests corrected. IEC 60891 temperature and irradiance corrections; Chinese reports citing NB/T 32034 or GB/T 39857 checked against the named text.
9. Challenge the model's rate. Non-linear field behaviour versus the linear data-sheet assumption.

Scope limits

This guide reads IEC webstore abstracts (texts not read), two NREL reports, the RdTools documentation, two Chinese standard records (texts not read) and four IEA PVPS Task 13 documents (one a fact sheet whose full report was not read). It quotes no plant data, no manufacturer warranty curve, no Chinese plant statistic and no clause of GB/T 39857-2021 or NB/T 32034-2016; the 2016 and 2018 NREL degradation papers could not be retrieved and are not cited.