How NREL Pushed the Limits of Accuracy in PV Performance Measurements

April 16, 2019 by By Dean Levi

When considering a new undertaking, I often ask, "Why?"
Why should we do this? How long will it take? What will we gain?

A researcher places a solar cell on an outdoor rack.

An NREL researcher prepares a solar module (in this case, a thin film module) for outdoor calibration. Using sunlight to calibrate primary reference cells helped NREL increase the accuracy of its PV performance measurements.

Photovoltaic (PV) manufacturers want reduced uncertainty in PV performance measurements because it allows them to improve the margin of return on the modules they sell. Investors and financiers want reduced uncertainty because it reduces the risk in their investments. Test labs studying module reliability want reduced uncertainty because it allows them to more accurately resolve degradation rates of PV modules. Researchers developing the next generation of PV cells want the highest accuracy possible in discriminating new world record devices.

Better PV performance measurements can have a big impact. So how can we reduce the uncertainty in PV performance measurements?

It All Starts with the Sun

In 2015, NREL began to push the limits of reducing uncertainty with fundamental enhancements to the way it calibrates primary reference cells. Primary reference cells are used by calibration labs to set the light intensity of the solar simulator. So more accurate calibration of primary reference cells provides increased accuracy for every type of PV device performance testing we do.

These fundamental enhancements to our primary reference cell calibration process (performed outdoors using sunlight) included better collimation, better stray light reduction, a new wider-range spectroradiometer, spectral fitting up to 10 um instead of 4 um, an enhanced spectral model, collection of many more data points, and temperature-dependent spectral response corrections (in place of linear temperature coefficients).

The combination of all of these improvements resulted in a six-fold decrease in the standard deviation and a reduction in uncertainty (k=2 coverage factor) from ±0.91% to ±0.40%.

A graph of calibration values from 1990 to 2016 in which the values are spread broadly for most of this timeframe but restrict into a much tighter cluster in 2016.

Control chart showing every calibration data point for NREL mono-silicon reference cell S01 starting in 1990. Each point corresponds to one spectral irradiance measurement. Red points were measured in 2016 with the hardware and software improvements. Calibration values are normalized to the mean 2016 value. The dashed lines are ±2 standard deviations.

This is a radical reduction of 56% from our previous uncertainty, and it has had major impacts on the rest of our PV performance measurement uncertainties. Prior to this, our uncertainty for secondary reference cells was ±1.6%. This new uncertainty for primary reference cells—plus new procedures to better correct for solar simulator spatial nonuniformity and for temperature correction of the reference cell spectral response—resulted in a new accredited uncertainty for the calibration of secondary reference cells of ±0.6%.

Although our accreditation strictly applies only to packaged reference cells, we achieve uncertainties very close to this for many other single junction PV devices, such as full-sized silicon wafer cells and many small-area research cells.

Better Reference Cells Improve Module Measurements

While these improvements for primary and secondary reference cells are impressive, the PV industry really cares about full-sized PV modules. For such large samples, the spatial nonuniformity of irradiance in the solar simulator dominates the uncertainty.

The reason spatial nonuniformity is such a big factor is the size mismatch between a PV module and a primary reference cell. A reference cell's area is typically 4 cm2, whereas modules are typically 16,000–20,000 cm2. Because of this, any nonuniformity in irradiance must be directly included in the uncertainty calculation.

Because our workhorse solar simulator for module calibration has ±3% spatial nonuniformity, our uncertainty for PV modules was ±3.3%. Without getting into all of the details (read more in our January 2018 blog on the module self-reference procedure), we figured out a way to defeat this spatial nonuniformity by calibrating the module's current outdoors using natural sunlight (which is very uniform). We then used the module itself in place of the reference cell to set the simulator irradiance level. This completely eliminates the effect of irradiance nonuniformity and enables us to reach an uncertainty of ±1.1% for module power.

These three uncertainties—±0.4% for primary reference cells, ±0.6% for secondary reference cells, and ±1.1% for modules—make NREL the most accurate laboratory in the world for PV performance calibration. Learn more and access our calibration services.