0
Technical Brief

# Solar Variability DataloggerOPEN ACCESS

[+] Author and Article Information
Matthew Lave

Sandia National Laboratories,
Livermore, CA 94550
e-mail: mlave@sandia.gov

Joshua Stein

Sandia National Laboratories,
Albuquerque, NM 87185

Ryan Smith

Pordis, LLC,
Austin, TX 78729

Contributed by the Solar Energy Division of ASME for publication in the JOURNAL OF SOLAR ENERGY ENGINEERING: INCLUDING WIND ENERGY AND BUILDING ENERGY CONSERVATION. Manuscript received December 17, 2015; final manuscript received June 14, 2016; published online July 28, 2016. Assoc. Editor: Philippe Blanc.The United States Government retains, and by accepting the article for publication, the publisher acknowledges that the United States Government retains, a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for United States government purposes.

J. Sol. Energy Eng 138(5), 054503 (Jul 28, 2016) (8 pages) Paper No: SOL-15-1432; doi: 10.1115/1.4034071 History: Received December 17, 2015; Revised June 14, 2016

## Abstract

To address the lack of knowledge of local solar variability, we have developed and deployed a low-cost solar variability datalogger (SVD). While most currently used solar irradiance sensors are expensive pyranometers with high accuracy (relevant for annual energy estimates), low-cost sensors display similar precision (relevant for solar variability) as high-cost pyranometers, even if they are not as accurate. In this work, we present evaluation of various low-cost irradiance sensor types, describe the SVD, and present validation and comparison of the SVD collected data. The low cost and ease of use of the SVD will enable a greater understanding of local solar variability, which will reduce developer and utility uncertainty about the impact of solar photovoltaic (PV) installations and thus will encourage greater penetrations of solar energy.

<>

## Introduction

High-frequency solar irradiance is rarely measured, and, hence, local short-term solar variability is rarely known. But knowledge of local solar variability is essential for grid integration studies which determine the impact of solar PV to electric grids. For example, solar variability leads to voltage fluctuations on distribution feeders; the magnitude and frequency of the voltage fluctuations depend on the amount of solar variability [1]. Measurements are scarce partially due to the high cost of measurement: current pyranometer plus datalogger pairs typically exceed $1000 [2], with additional costs for power and communications and complicated setup. These pyranometers were designed for high accuracy (relevant for annual energy estimates). For solar variability measurements that capture short-term changes in irradiance, though, precision is more important than accuracy, suggesting that low-cost irradiance sensors with high precision can have similar ability to measure solar variability as pyranometers even if they are not as accurate for annual energy measurements. A handful of previous projects have attempted to design low-cost solar irradiance sensors, such as Mancilla-David et al. [3], Cruz-Colon et al. [4], and The Navy Research Lab [5], but they all focused on energy monitoring of PV system performance and not on solar variability analysis. Some groups have monitored solar variability with conventional pyranometers, including the National Renewable Energy Laboratory (NREL) [6], the Hawaii Electric Company (HECO) [7], the Sacramento Municipal Utility District (SMUD) [8], and the Electric Power Research Institute (EPRI) [9], but the geographic coverage of deployed sensors was limited to a few select areas, and the data were generally not available to other researchers (with the notable exception of the NREL Oahu data).1 In this work, we describe a solar variability datalogger (SVD) that uses a low-cost PV cell to measure irradiance and has integrated power, data storage, and communications, resulting in simple installation and operation. This will enable ubiquitous deployment and hence a greater understanding of local solar variability. ## Variability Measurement With Low-Cost Sensor Five low-cost test sensors designs were evaluated for their ability to measure solar variability: two PV cell based designs and three photodiode-based designs. PV01 and PV02 were the PV-based designs: PV01 has a glass window mounted under the sensor casing and PV02 has the glass window mounted above the sensor casing. The photodiode-based designs were called PD01, PD02, and PD03. PD01 had a Teflon diffuser 4.13 mm from the photodiode, PD02 had a clear FEP diffuser 4.13 mm from the photodiode, and PD03 had a Teflon diffuser 1.4 mm from the photodiode. Each of the five test sensors plus a LICOR LI-200SL was deployed in Austin, TX. Figure 1 shows the test setup. Evaluation of each test sensor against the reference LICOR was conducted over 7 hrs on a highly variable day (13:00–20:00 on Apr. 6, 2015). Accelerated project timelines limited testing to only this 1 day. While a longer test period would have given a more definitive comparison among sensor designs, this short test still assists in down-selection of well-performing designs. Section 4.2 contains validation over a longer time period for the selected sensor. Figure 2 shows the histograms of differences in measured irradiance between the test devices and the LICOR. Both the PV01 and PV02 devices have distributions that are centered close to zero, though the PV01 device has a bimodal distribution with a peak around 0 W m−2 caused by measurements within 1 hr of solar noon and another peak around 10 W m−2 caused by measurements later in the day. The PD01 and PD03 devices almost always measured less irradiance than the LICOR (negative values in Fig. 2), perhaps explained by the different diffuser materials used in the PD01 and PD03 devices versus the LICOR. The PD02 device shows significant differences (there is a large peak at the negative boundary of the histogram), which were caused by the diffuser casing shading the photodiode at low sun angles: the clear FEP diffuser material was not strong enough to eliminate this shading. While matching irradiance measurements is important (e.g., PD02 is eliminated from consideration due to its significant shading at low sun angles), the goal is to accurately measure solar variability. Ramp rates (RRs) are the changes in solar irradiance over a certain time interval Display Formula (1)$RRΔt(t)=1Δt(∑tt+ΔtIrr−∑t−ΔttIrr)$ where $RRΔt(t)$ is the irradiance RR for timescale $Δt$ at time t. Figure 3 compares distributions of 30 s ($Δt=30s$) RRs measured by the LICOR to the 30 s RRs measured by the test devices. These distributions describe the solar variability: distributions shifted up and to the right (higher probabilities (y-axis) of large ramps (x-axis)) indicate larger amounts of variability. To directly quantify solar variability, we used the variability score (VS) [1]. VSs are small (0–10) for clear conditions with low variability and large (>100) for highly variable conditions. The VS is the maximum value of the quantity RR magnitude (RR0, expressed as a percentage of 1000 W m−2) times RR probability, multiplied by 100 to give an easier number to interpret. Display Formula (2)$VS=100×max[RR0×P(|RR|>RR0)]$ For example, the LICOR RR distribution had the maximum value of RR magnitude times RR probability at 150 W m−2 ramp with 16.48% probability. Multiplying 15.0% (150 W m−2 relative to 1000 W m−2) by this 16.48% probability gives 2.47%, and scaling by 100 gives the VS of 247. The VS was used to quantify the differences in the RR distributions between the LICOR and the test devices. This VS comparison was chosen over other methods—such as the Kolmogorov–Smirnov test—which equally weight probability differences over all the RR magnitudes. For solar energy applications, small errors at large magnitude RRs are much more important than large errors at small ramp magnitudes. For example, estimating that only 1% of ramps are larger than 400 W m−2, when in truth 2% of ramps are larger than 400 Wm−2 could lead to undersizing of storage at a PV plant. Conversely, estimating that 10% of ramps are smaller than 10 W m−2, when in truth 20% of ramps are smaller than 10 W m−2 would likely have no impact on PV plant operation. Since the VS uses the quantity RR magnitude times probability, it more heavily weights the larger ramps. The calculated VSs are shown in the legends in Fig. 3, and the error (test device versus LICOR) in VS is listed as text in each plot. The very large VSs measured by both the LICOR (247) and the test devices (225–246) confirm that this was a highly variable day. Consistent with the irradiance comparison (Fig. 2), the PV cell devices both had smaller VS errors than the photodiode devices. ## SVD Description Using the results of Sec. 2, PV02 was determined to be the most promising design. A device was built based on the PV02 sensor design, and also including data storage, wireless communication, and battery power integrated into a single standalone unit. Because of these integrated features, we refer to the device as a “solar variability datalogger” (SVD) instead of just a variability sensor, to emphasize its ability to be a complete solution for monitoring solar variability. An overview of features is shown in Fig. 4, and further details are given in Secs. 3.1, 3.2, and 3.3. ###### Hardware. The outside of the SVD is a white and gray weatherproof casing measuring roughly 161 mm long × 80 mm wide × 58 mm tall (6.3 in. × 3.2 in. × 2.3 in.). As seen in Fig. 4, a heat-resistant borosilicate glass window exposes the monocrystalline silicon PV cell, and a serial data connection is accessible on the front of the device. Inside the casing are a circuit board, a global positioning system (GPS) module, a Wi-Fi module, an accelerometer, and a lithium metal battery. The casing is watertight, but desiccant is included inside to further protect against water damage. In total, all the hardware components cost approximately$280 at time of writing, though this is expected to be reduced to approximately $150 in mass production (1000 units). Labor costs for assembly will also be reduced in mass production. ###### Operation and Data Collection. On power-up, the SVD uses the GPS to obtain the current time to sync the on-board clock and the location to compute sunrise and sunset times. In normal operation, the SVD samples the irradiance once every second during daylight hours and hibernates at night. Differences in subsequent 1 s measurements are recorded in RR histogram bins which range from −500 to 500 W m−2 in 2 W m−2 increments. Additionally, 30 s averages of the 1 s irradiance samples are computed and recorded. The SVD was designed to conserve power and minimize recorded data sizes. Irradiance sampling takes only about 100 ms, so the rest of the time (90% of the time) the PV cell is used to charge a supercapacitor attached to the circuit board, which can be discharged as needed to save battery energy. To limit data sizes, only 30 s averages and RR histograms are recorded. On a 12-hr day, for example, this reduces the amount of recorded data by a factor of 22 (1440 30 s averages plus 500 histogram bins compared to 43,200 1 s samples). This not only reduces the amount of required on-board memory but also reduces data transmission times (Sec. 3.3). The SVD has enough data storage for up to 140 days without overwriting. Two special operation modes can be activated by holding a magnet to an indicated spot on the device (seen in Fig. 4 on the left of the SVD): (1) leveling mode uses the accelerometer to determine the device's orientation and helps the user to level the device by blinking the light-emitting diodes installed around the PV cell in the uneven direction(s), and (2) data dump mode prints all the data stored on the device to the serial output (Sec. 3.3). ###### Data Transmission. There are three different ways to retrieve data from the SVD: serial cable, Wi-Fi, and cellular modem. Data download via serial cable is the simplest and most reliable way to download SVD data. A serial cable is connected to the SVD, with the other end connected via universal serial bus (USB) to a computer. A serial data-logging program, such as putty,2 is used to log the output from the SVD. A 1 s irradiance measurement is output in real time (i.e., every second). This allows for collection of 1 s data if a serial cable is continuously connected to the SVD. If a cable is not continuously connected and instead there are periodic visits to the SVD to collect data, a dump of all the data stored on the device is activated by holding a magnet to the SVD while a serial cable is connected. The second data transmission option is through Wi-Fi. When Wi-Fi is setup, the SVD connects to a web server at sunset and uploads the 30 s averages and RRs histogram from the current day. This data is then hosted on the web server for later retrieval and data analysis. Finally, the SVD can also upload its data via cellular modem. This requires a separate enclosure with its own battery and solar panel since the cellular modem requires more power than the SVD, increasing the cost but enabling installation in a remote location. The SVD controls power to the cellular modem such that the modem is only on for approximately 2 min at sunset while that day's data are uploaded. Data upload protocol with the cellular modem is identical to upload using Wi-Fi. The data are uploaded to the same web server used for Wi-Fi uploads. Even if Wi-Fi or cellular modem is used as the main data transmission method, serial download can always be used as a backup to retrieve all the historic data. This ensures that no data are lost. ## Field Testing of SVD The SVD was installed at three different locations: Albuquerque, NM; Livermore, CA; and Austin, TX. The Albuquerque and Livermore installations are co-located with a pyranometer measuring global horizontal irradiance (GHI) (Kipp and Zonen CM21 in Albuquerque and Eppley PSP in Livermore), allowing for validation of measurements. These thermopile pyranometers have slower response times (5 s, see Refs. [10] and [11]) than the SVD, but are still appropriate for 30 s variability comparisons. Figure 5 shows the Livermore installation and co-located weather station with GHI measurement. At the time of writing, data have been collected in Albuquerque for approximately 7 months, and in Livermore and Austin for approximately 5 months. ###### Clear-Sky Calibration. In both Albuquerque and Livermore, SVD GHI measurements were found to be low compared to the pyranometers. To account for this, we calibrated the SVD using a clear-sky model based on the method described in Sec. 4 of Reno and Hansen [12]. We used the clear-sky model described by Ineichen and Perez [13], as implemented in the PVLib Toolbox [14], which requires only location and timestamps to compute a clear-sky model. Thus, the clear-sky calibration can be applied to any SVD regardless of location. The calibration coefficients were found for the three locations based on all the available data: the Albuquerque SVD calibration factor was 1.1893 (i.e., the SVD data were about 19% low compared to the clear-sky model), the Livermore calibration was 1.1956, and the Austin calibration was 1.1902. The consistency of calibration factors among the different devices in different locations may indicate a universal calibration factor of approximately 1.19. Figure 6 compares both the raw and clear-sky calibrated SVD GHI measurements in Albuquerque to the co-located CMP21 on three clear days. Subsequent evaluation revealed an (undocumented) change in the encapsulant used in the PV cell between initial testing—when the original calibration factor was determined based on the short circuit current—and the final device assembly. This likely explains the $∼19%$ low SVD measurements and a simple adjustment to the calibration factor should fix the issue. However, the clear-sky calibration is still valuable as it can be used to mitigate any other biases that may exist in SVD measurements (e.g., soiling). ###### Validation of Variability Measurements. Variability measurements from the SVD were compared to pyranometer measurements in Albuquerque and Livermore. Figure 7 shows both the RR distributions over all test days (left plots), and the day-by-day VSs (right plots) for both the SVD and the pyranometer. The RR distributions over all days show that the SVD closely matched the pyranometer-measured variability, as seen by comparing the VSs (shown in the legend of the left plots). In Albuquerque, the difference between the SVD and the pyranometer VSs was 1.8%, and in Livermore the difference was 3.2%. The two SVDs capture the different VSs for the Albuquerque and Livermore data samples (Livermore less variable than Albuquerque), consistent with the pyranometer variability measurements. Similarly, the scatter plots show day-by-day VSs that are closely matched between the SVDs and the pyranometers. The day in Albuquerque with the greatest difference between SVD and pyranometer was Dec. 29, 2015 (SVD VS = 47 and CMP21 VS = 83), a day which was preceded by 3 days of snowfall and freezing temperatures. It appears that different rates of snow melt led to the discrepancy in VSs. The difference between Albuquerque and Livermore solar variability is also seen in the scatter plots: most days in Albuquerque had VSs less than 100, with a few days when VSs exceeded 150, while most days in Livermore had VSs less than 50, with no days exceeding 150. ###### Comparison Between Locations. A main reason for developing the SVD was to enable measurements at various locations to allow for comparison of solar variability. The three SVD measurement locations are compared directly in Fig. 8, which shows a calendar plot of SVD GHI measurements for the month of November, 2015. GHI values are plotted against coordinated universal time to show the impact of the sun's movement through the sky and to separate the curves for better readability. Gaps in the data presented in Fig. 8 are predominantly due to research activities such as removing the device to install software or hardware updates. Also included in Fig. 8 are the 30 s VSs for each day to allow for quantitative comparisons of daily solar variability. The data presented in Fig. 8 highlight how much solar variability can change by day and by location, showing the importance of measuring the local solar variability. On different days at the same location, the VSs range from over 100 (partly cloudy days) to nearly zero (clear days). Similar variation can be seen at different locations on the same day. For example, on November 10: Albuquerque was clear; Livermore had high variability, especially in the afternoon, likely due to partly cloudy conditions; and Austin had low variability all day, likely due to persistent, only slightly varying cloud cover. The VSs for November 10 quantify the low variability in Albuquerque and Austin and high variability in Livermore: Albuquerque VS = 6, Livermore VS = 114, and Austin VS = 16. ## Conclusion We developed and deployed a low-cost SVD to address the lack of high-frequency solar variability measurements which are essential for accurate modeling of the impact of solar (PV) to electric grids. First, we tested low-cost irradiance sensors (PV cells and photodiodes) as a proof-of-concept that low-cost sensors could match solar variability measurements of high-cost pyranometers. Then, based on the most promising sensor (a PV cell), we developed the SVD, which is an integrated solution for measuring solar variability: irradiance sensor, data logging, power, GPS, and communications are all included in the SVD. The SVD hardware is expected to cost approximately$150 in mass production, though labor costs for assembly will increase the unit cost somewhat. Data collected from SVD units were both validated against pyranometers and compared to one another to show differences in solar variability by day and by location.

The low cost and ease of installation and operation of the SVD will enable ubiquitous deployment and hence a greater understanding of local solar variability. For most applications, the SVD measured solar variability should be combined with models which account for the smoothing due to spatial diversity of PV modules across a PV plant or distribution feeder. Such models are described in Ref. [15], and their performance is evaluated in Ref. [16]. Alternatively, a fleet of SVD units could be deployed across a PV plant or distribution feeder to directly measure the spatial diversity.

We see many potential applications for the SVD. Solar developers could install one or more SVDs at prospective PV plants to understand how variable the PV output will be—this will be especially important for estimating storage capacities required to counter solar variability in locations (e.g., Puerto Rico [17]) with RR restrictions. Utilities could install SVDs on their distribution feeders to understand the impact of the local solar variability to their distribution grid operations (e.g., as demonstrated in Ref. [1]), thus informing integration costs and maintenance schedules. Researchers could install SVDs to monitor solar variability at many more locations than currently have measurements. Overall, the SVD will reduce uncertainty about the impact of solar PV installations and will encourage greater penetrations of solar energy, where appropriate.

## Acknowledgements

Sandia National Laboratories is a multiprogram laboratory managed and operated by the Sandia Corporation, a wholly owned subsidiary of the Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under Contract No. DE-AC04-94AL85000. This work was supported by the U.S. Department of Energy, SunShot Initiative, under Award No. SNL 29096. Report No. SAND2015-10902 J.

## References

Lave, M. , Reno, M. J. , and Broderick, R. J. , 2015, “ Characterizing Local High-Frequency Solar Variability and Its Impact to Distribution Studies,” Sol. Energy, 118, pp. 327–337.
Srikrishnan, V. , 2015, “ Estimation of Direct Normal Irradiance With Multi-Pyranometer Arrays and Artificial Neural Networks,” Master's thesis, The Pennsylvania State University, State College, PA.
Mancilla-David, F. , Riganti-Fulginei, F. , Laudani, A. , and Salvini, A. , 2014, “ A Neural Network-Based Low-Cost Solar Irradiance Sensor,” IEEE Trans. Instrum. Meas., 63(3), pp. 583–591.
Cruz-Colon, J. , Martinez-Mitjans, L. , and Ortiz-Rivera, E. , 2012, “ Design of a Low Cost Irradiance Meter Using a Photovoltaic Panel,” 38th IEEE Photovoltaic Specialists Conference (PVSC), Austin, TX, June 3–8, pp. 002911–002915.
Parry, D. , 2013, “ NRL Develops Low Cost, High Efficiency Solar Sensor,” U.S. Naval Research Laboratory, Washington, DC.
Sengupta, M. , and Andreas, A. , 2010, “ Oahu Solar Measurement Grid (1-Year Archive): 1-Second Solar Irradiance; Oahu, Hawaii (Data),” National Renewable Energy Laboratory, Golden, CO, Technical Report No. DA-5500-56506.
Dangelmaier, L. , 2012, “ HECO Companies Experience With Distributed PV,” PJM/EPRI/NREL Inverter Based Generation Interconnection Workshop Proceedings.
Sacramento Municipal Utility District, 2014, “ High Penetration Photovoltaic Initiative,” Sacramento Municipal Utility District, Sacramento, CA.
EPRI, 2016, “ Distributed PV Monitoring and Feeder Analysis,” Electric Power Research Institute, Palo Alto, CA.
Kipp and Zonen, 2015, “ CMP21 Instruction Manual,” Kipp and Zonen, Delft, The Netherlands.
The Eppley Laboratory, 2013, “ Standard Precision Pyranometer, Model SPP,” The Eppley Laboratory, Newport, RI.
Reno, M. J. , and Hansen, C. W. , 2016, “ Identification of Periods of Clear Sky Irradiance in Time Series of {GHI} Measurements,” Renewable Energy, 90, pp. 520–531.
Ineichen, P. , and Perez, R. , 2002, “ A New Airmass Independent Formulation for the Linke Turbidity Coefficient,” Sol. Energy, 73(3), pp. 151–157.
SNL, 2014, “ PVLib Toolbox: Ineichen Clear Sky Model,” Sandia National Laboratory, Albuquerque, NM.
Lave, M. , and Kleissl, J. , 2013, “ Cloud Speed Impact on Solar Variability Scaling: Application to the Wavelet Variability Model,” Sol. Energy, 91, pp. 11–21.
Lave, M. , 2012, “ Comparison of Errors in Solar Power Plant Variability Simulation Methods,” Sandia National Laboratories, Albuquerque, NM, Technical Report No. SAND2015-0156.
Gevorgian, V. , and Booth, S. , 2013, “ Review of PREPA Technical Requirements for Interconnecting Wind and Solar Generation,” National Renewable Energy Laboratory, National Renewable Energy Laboratory, Golden, CO, Technical Report No. NREL/TP-5D00-57089.
View article in PDF format.

## References

Lave, M. , Reno, M. J. , and Broderick, R. J. , 2015, “ Characterizing Local High-Frequency Solar Variability and Its Impact to Distribution Studies,” Sol. Energy, 118, pp. 327–337.
Srikrishnan, V. , 2015, “ Estimation of Direct Normal Irradiance With Multi-Pyranometer Arrays and Artificial Neural Networks,” Master's thesis, The Pennsylvania State University, State College, PA.
Mancilla-David, F. , Riganti-Fulginei, F. , Laudani, A. , and Salvini, A. , 2014, “ A Neural Network-Based Low-Cost Solar Irradiance Sensor,” IEEE Trans. Instrum. Meas., 63(3), pp. 583–591.
Cruz-Colon, J. , Martinez-Mitjans, L. , and Ortiz-Rivera, E. , 2012, “ Design of a Low Cost Irradiance Meter Using a Photovoltaic Panel,” 38th IEEE Photovoltaic Specialists Conference (PVSC), Austin, TX, June 3–8, pp. 002911–002915.
Parry, D. , 2013, “ NRL Develops Low Cost, High Efficiency Solar Sensor,” U.S. Naval Research Laboratory, Washington, DC.
Sengupta, M. , and Andreas, A. , 2010, “ Oahu Solar Measurement Grid (1-Year Archive): 1-Second Solar Irradiance; Oahu, Hawaii (Data),” National Renewable Energy Laboratory, Golden, CO, Technical Report No. DA-5500-56506.
Dangelmaier, L. , 2012, “ HECO Companies Experience With Distributed PV,” PJM/EPRI/NREL Inverter Based Generation Interconnection Workshop Proceedings.
Sacramento Municipal Utility District, 2014, “ High Penetration Photovoltaic Initiative,” Sacramento Municipal Utility District, Sacramento, CA.
EPRI, 2016, “ Distributed PV Monitoring and Feeder Analysis,” Electric Power Research Institute, Palo Alto, CA.
Kipp and Zonen, 2015, “ CMP21 Instruction Manual,” Kipp and Zonen, Delft, The Netherlands.
The Eppley Laboratory, 2013, “ Standard Precision Pyranometer, Model SPP,” The Eppley Laboratory, Newport, RI.
Reno, M. J. , and Hansen, C. W. , 2016, “ Identification of Periods of Clear Sky Irradiance in Time Series of {GHI} Measurements,” Renewable Energy, 90, pp. 520–531.
Ineichen, P. , and Perez, R. , 2002, “ A New Airmass Independent Formulation for the Linke Turbidity Coefficient,” Sol. Energy, 73(3), pp. 151–157.
SNL, 2014, “ PVLib Toolbox: Ineichen Clear Sky Model,” Sandia National Laboratory, Albuquerque, NM.
Lave, M. , and Kleissl, J. , 2013, “ Cloud Speed Impact on Solar Variability Scaling: Application to the Wavelet Variability Model,” Sol. Energy, 91, pp. 11–21.
Lave, M. , 2012, “ Comparison of Errors in Solar Power Plant Variability Simulation Methods,” Sandia National Laboratories, Albuquerque, NM, Technical Report No. SAND2015-0156.
Gevorgian, V. , and Booth, S. , 2013, “ Review of PREPA Technical Requirements for Interconnecting Wind and Solar Generation,” National Renewable Energy Laboratory, National Renewable Energy Laboratory, Golden, CO, Technical Report No. NREL/TP-5D00-57089.

## Figures

Fig. 1

Test setup with (from left to right) PV01, PV02, LICOR pyranometer, PD01, PD02, and PD03

Fig. 2

Histograms of differences in irradiance measurements between the test sensors and the LICOR over 7 hrs on a highly variable day. Included in the top right of each plot are the mean bias difference (MBD) and root-mean squared difference (RMSD) between the test devices and the LICOR (negative MBD values indicate that the test device measured less irradiance than the LICOR).

Fig. 3

A 30 s RR distribution for LICOR and test sensors over 7 hrs on a highly variable day

Fig. 4

SVD features

Fig. 5

SVD (circled) installed at a weather station in Livermore, CA. The instrument to the top left of the solar panels is the PSP GHI measurement.

Fig. 6

GHI time series measured in Albuquerque on three clear days by a CMP21 and an SVD. The SVD is shown without and with clear-sky correction.

Fig. 7

(Left) A 30 s RR distribution collected over all days in the measurement period for pyranometers (solid lines) and SVDs (dashed lines), including VSs. The zoomed-in y-axis range of 0–10% probability is shown in larger plot, with the full 0–100% range shown in the inset plot. (Right) Scatter plot of daily VSs (one VS per day in measurement period) comparing SVD to pyranometers.

Fig. 8

(Top) Calendar plot of SVD GHI measurements at different locations in November 2015. (Bottom) Comparison of 30 s VSs for each day.

## Discussions

Some tools below are only available to our subscribers or users with an online account.

### Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related Proceedings Articles
Related eBook Content
Topic Collections