With the deepening of people’s understanding of solar energy resources and the increasing demand for solar energy resource information, the National Renewable Energy Laboratory (NREL) has updated and improved the national solar radiation database (nsrdb). The database was first updated in the mid-1990s, covering the solar radiation data and meteorological data of 239 locations from 1961 to 1990 (the 1995 version of the user manual of nsrdb database 1961-1990). Although some sites have a certain amount of irradiance measurement data, the nsrdb database mainly contains the modeling values determined by meistat model (Maxwell, April 1998).
Meistat uses cloud cover data, aerosol data and other meteorological data to calculate the incident GHI and DNI values. Statistically, the incident GHI and DNI values are similar to the actual measured hourly irradiance data. There are some minor problems in metstat model, which affect the irradiance evaluation in cloudy periods (Vignola, April 1997). However, the irradiance data set generated by the model is statistically matched with the measured irradiance data.
The nsrdb database reanalyzes part of the solmet / ersatz data and stores it in the updated database. Considering that when reporting meteorological mean, it is usually sufficient to use a database covering 30 years of meteorological data, so there is no need to re analyze the data before 1961. The measured data provided by some solmet stations have also been reanalyzed and converted into local time, but the analysis requires a lot of time investment, and the quality of the data is affected by the instrument and its calibration to some extent. Therefore, many solmet stations do not re digitize and adjust the chart data, but use the meteorological data to establish the solar energy resource model.
The US West Coast site in the solmet dataset contains most of the modeling data. Some solmet / ersatz observation stations are not included in the updated nsrdb database because they no longer exist or do not have sufficient recorded data.
When all meteorological data and aerosol data are available, the monthly mean metstat data has an uncertainty of 9% at the 95% confidence level. In some time periods, the meteorological values in the nsrdb database must be temporarily inferred from the existing data, and these time periods have higher uncertainty. The largest uncertainty in the nsrdb database is about 24% and will appear at some sites. The records of these sites come from different time periods to fill in the blank records.
Many sites in the nsrdb database from 1991 to 2005 contain the change cycle. Tracking the GHI values of Phoenix, Arizona, dagate and California from 1961 to 1994, the results are quite consistent. However, between 1995 and 1997, there was a sudden huge difference between the GHI of Fenghuang city and dagate. The detection of uncertain data shows that the difference of GHI between the two places has increased significantly, indicating that there is still a gap between documents filled with data from other periods. Time periods marked with more uncertainty will cause data deviation, so they cannot be used to create evaluation data sets.
The statistical properties of the metstat model and the assumptions used to generate the aerosols used in the model mask the long-term climate change trend in the irradiance data. In addition, the systematic error caused by the uncertainty in the meteorological data and the systematic error in the irradiance data used to verify the model also mask this trend. The uncertainty range of GHI value obtained by metstat model is ± 9%. This means that the true average GH field obtained from several measurements under similar conditions must be within the range of 9% of the modeled value 95% of the time (nsrdb database (Updated) user manual 1991-2005, published in April 2007).
In 2007, NREL expanded the nsrdb database using meteorological data and satellite images to generate irradiance values for 1454 observation stations from 1991 to 2005. From 1998 to 2005, irradiance data came from satellite images and other meteorological data and auxiliary data (2007)
The model completed by the 1991-2005 nsrdb database (updated version) user manual released in April. Most of the data values from 1991 to 1994 come from the modified metstat model proposed by Vignola (Vignola, April 1997), so that better statistics can be generated during cloudy weather periods. From 1994 to 1998, irradiance values were derived from the modified metstat model. The model does not use the percentage of cloud amount observed by the naked eye before metstat model, but uses the cloud height data and other data of ground automatic meteorological observation station (ASOs / AWOS). ASOs / AWOS is an automatic meteorological observation station located at or near the airport.
In the transition from human observation to automatic meteorological observation station, incomplete or missing records sometimes occur. In case of missing records, in order to maintain the continuity and integrity of the data set, other data of similar period (up to 1 year) can be used for replacement. Before and after the missing data, use meteorological data to select relevant data to fill the missing part. The more missing parts, the less reliable this method is. The data generated by using the input meteorological data of other periods have high uncertainty, with some uncertainty values of GHI of 24% and some uncertainty values of DNI of 27%. The data with high uncertainty is not enough to predict the efficiency of large-scale solar projects, and may cause deviation in the results. Therefore, it is necessary to constantly check the uncertainty related to each data point and exclude the data generated during the period of high uncertainty.
Since 1998, satellite irradiance values have been available at all stations in the nsrdb database, and the relevant records are very complete. Therefore, satellite irradiance data were used for all locations in the nsrdb database from 1998 to 2005. NREL recently updated the data files of 2010, and these files can also be used in the nsrdb database from 1998 to 2005 (including metstat modeling data and some high-quality measured data used at ASOs / AWOS meteorological observation stations or nearby stations).
The satellite data provided by the State University of New York (SUNY) at Albany comes from hourly satellite images (Vignola and Perez, 2004), and the images come from the visible light channel on the goes meteorological satellite.
SUNY satellite data is located on a 1 ° grid with a grid range of about 10km. The goes West satellite image and SUNY data set are the contents of a certain whole point time, representing the average irradiance 0.5h before and after the current time.
The goes East satellite image is the image content 15 minutes after the current time, representing the data 15 minutes before and 45 minutes after the current time. (the goes satellite now generates images every half an hour.) SUNY Albany model adopts grid aerosol data (developed by NREL for nsrdb database) to assist data modeling. In order to merge the satellite data into the nsrdb database, the satellite image must be translated for 0.5h or 15min to be consistent with the meteorological data in the data set. The user manual of nsrdb database introduces the averaging process in detail (user manual of nsrdb database (updated version) from 1991 to 2005, released in April 2007). The increase of time translation increases the random uncertainty in the data by 1% ~ 2%. At 95% confidence level, the uncertainty of monthly average daily GHI presented by satellite images is 8%, while that of DNI is 15%. This once again shows that the average of several measured values of GHI or DNI under similar conditions is uncertain. The user manual details the uncertainties in the data of nsrdb database (1961-1990 nsrdb database user manual of 1995 version and 1991-2005 nsrdb database (updated version) user manual of April 2007). This paper introduces the uncertainty of satellite data in section 5.4.
For research purposes, the official website of NREL provides solar radiation data files for the nsrdb database from 1991 to 2005. The website contains metstat modeling data, ground measured irradiance data and satellite irradiance data (available since 1998). Documents containing metstat model generated values and satellite values can be used to illustrate the uncertainty of irradiance values. It is found that the metstat value and satellite value are the same as the measured value, but there are satellite values at all locations throughout the country, so the satellite irradiance data is selected as the basis.
With the continuous improvement of the model that can export satellite irradiance values, new satellite data sets have been released. Most of the new satellite irradiance values are within the uncertainty range of the original data set.