Solar Resource Monitoring Station

Data Documentation

 

 

 

King Abdullah City for Atomic and Renewable Energy (K.A.CARE) is collecting a one-of-a-kind, high-quality dataset of solar resources at over 35 monitoring sites throughout Saudi Arabia.  At every step of the way since network planning began in 2011, K.A.CARE has worked with world experts to ensure proper station design, equipment installation, measurement validation, instrument calibration, and station maintenance. These experts include scientists and engineers from the U.S. National Renewable Energy Laboratory and Battelle, along with an international advisory committee with representatives from leading institutions within and outside of Saudi Arabia such as the International Renewable Energy Agency (IRENA).  This report documents the data collection, quality review, aggregation, and uncertainty calculation processes.  Conservative uncertainty values are provided for each data point, which can be accessed in the data download section of the Renewable Resource Atlas (https://rratlas.kacare.gov.sa).  Additional information on station equipment and parameters is available from the Atlas under Reports (see the Solar Monitoring Network Summary Report and Individual Station Reports).  

Data Collection and Quality Review Process

Solar resource data is collected at 1-min resolution at all stations.  Each station includes redundant radiation sensors for quality assurance.  K.A.CARE data systems provide live data monitoring capabilities, such that field technicians can be quickly dispatched to investigate any problems.  Field technicians also perform station cleaning and maintenance according to specific protocols, with careful documentation, as follows:

§  Twice weekly for Tier 2 (Rotating Shadowband Radiometer, RSR) stations

§  Five times per week for Tier 1 (Research Grade, thermopile radiometer) stations

Both automatic and manual data processes are employed for daily data quality review.  A series of checks and tools such as SERI-QC[1] are implemented by K.A.CARE staff.  Short durations of data anomalies that occur during verified cleaning periods are filled with values through interpolation or other methods.  Any remaining data gaps of more than 1.5 hours are flagged as missing data. Individual station reports list the percentage of time that the station has been online since inception.

Data Aggregation and Uncertainty                           

For user convenience, K.A.CARE aggregates the 1-min solar resource values as summarized below.

Aggregation

Units

Description

Notes

5, 10, 15, and 30-min

W/m2

Averages of 1-min data

Only available to K.A.CARE staff; 1-min data include source and quality flags

Hourly

Wh/m2

Integration of 1-min data

Aggregation for 4:00 p.m. represents measurements taken from 3:01 to 4:00 p.m.

Daily

Wh/m2/ day

Daily Total

Daily aggregation represents the interval 00:01 to 24:00

Monthly

Wh/m2

Average Daily Total Across the Month

Represents all days in month

Annual

Wh/m2

Average Daily Total Across the Year

For complete calendar years of data only (Jan 1 – Dec 31)

Aggregated values are NOT calculated for periods where more than 20% of measured data is missing (e.g., due to equipment malfunction) for a given station and aggregation period.  In data downloads, missing values are reported as -9900.

Next, the uncertainty of each data point is calculated to reflect the nominal uncertainty of the monitoring equipment, the interval since cleaning, and data quality review flags (accounting for any filling and solar component comparisons).  Uncertainty calculations are based on a simplified use of the Guide to the Expression of Uncertainty (GUM) Method.[2] The table below lists the base nominal uncertainty values for 1-min values from monitoring equipment, drawn from experience and peer-reviewed scientific literature.

Monitoring Equipment

Parameter

% Base Nominal Uncertainty

Tier 1 (Research Grade) Stations

 

 

Thermopile pyranometer (un/shaded)

GHI, DHI

+/- 2%

Thermopile Pyrheliometer

DNI

+/- 2%

Tier 2 (Mid-Level) Stations

 

 

Rotating Shadowband Radiometer

GHI, DHI, DNI*

+/- 5%

*Note that DNI is calculated rather than measured for Tier 2 (RSR) stations.

Additional uncertainty is included for each day since cleaning (0.25% per day for Tier 2 stations and 0.45% per day for Tier 1 stations), and data quality review flags, corresponding with uncertainty levels up to 20%, are added if values are outside empirical limits.  Uncertainty levels for each 1-min data point are used to calculate the standard uncertainty assuming a rectangular distribution of uncertainties.

The 1-min uncertainty values are then aggregated to provide an uncertainty estimate for each data point reported in the Atlas.  The uncertainty reported in the Atlas represents a conservative U95 value, such that 95% of the data in a given aggregation (such as an hourly average of 1-min values) have lower uncertainty than the hourly uncertainty value reported.  Uncertainty is reported in measurement units, as a +/- value. Percent uncertainty can be calculated by dividing the reported uncertainty by the measured value. 

As of mid-2014, the U95 uncertainty levels achieved by the RRMM Network for monthly average daily totals are as follows:

·         Most GHI values fall into the 6-7% uncertainty (U95) category

·         Most DNI values fall into the 7-9% uncertainty (U95) category

The selection of the U95 method for calculating data uncertainty is intended to provide data users a very high degree of confidence that the data is as accurate as reported.  Data users need to be careful in comparing these uncertainty values to those reported for other sources of monitored and modeled data, since not all data providers use these conservative methods, and modeled data is typically compared to ground monitors which are treated as “truth” or 100% accurate.

Alternative methods which could be utilized by others to calculate uncertainty include reporting an “Average” uncertainty for a given aggregation period, rather than a U95 level.  Also, data anomalies during cleanings could be removed, rather than filled, which would result in a slightly less complete series of data with slightly lower (better) uncertainty.  K.A.CARE is currently exploring the impact of a missed cleaning, and whether the resulting “penalty” to data uncertainty is justified.  K.A.CARE is also exploring benchmarks of solar resource monitoring networks worldwide, and exploring alternate uncertainty calculation methods.

To request additional documentation, contact K.A.CARE at atlasinfo@energy.gov.sa.

 



[1] http://www.nrel.gov/docs/legosti/old/5608.pdf

[2] BIPM; IEC; IFCC; ISO; IUPAP; OIML. (1995). Guide to the Expression of Uncertainty in Measurement, ISO TAG4, Geneva.