Assessing the Accuracy of NREL's National Solar Radiation Data Base (NSRDB)

Since the early 1990s, the National Solar Radiation database (NSRDB) has served as a primary source of data for many solar energy system performance and economic models, has assisted in renewable energy project development and pre-feasibility studies, and supported solar energy system operations. Recognizing the importance of this foundational data to industry and other stakeholders, the U.S. Department of Energy has funded the development of the National Solar Radiation Database (NSRDB) by the National Renewable Energy Laboratory (NREL). The current NSRDB (1998-2015) uses a physics-based Physical Solar Model (PSM) to produce satellite-based gridded solar irradiance (global horizontal irradiance (GHI), direct normal irradiance (DNI), and diffuse horizontal irradiance) at a 4-km by 4-km spatial and half-hourly temporal resolution. Implementing a comprehensive uncertainty quantification approach for solar resource such as NSRDB is essential when discussing bankable data for all phases of solar energy conversion projects, from the conceptual phase to routine solar power plant operation. To ensure confidence in the product it is essential to understand the spatial and temporal accuracy of the 18 years of NSRDB data set using a standardized method of quantifying data uncertainty. Understanding the impacts of clouds and other meteorological constituents on the solar resource and quantifying intra-/inter-hour, seasonal, and inter-annual variability are essential for accurately designing utility-scale solar energy projects. Nine high quality ground-based measurement stations were selected to evaluate the NSRDB dataset. The comparisons of the gridded-NSRDB to the ground-based data demonstrated biases of +/- 5% for GHI and +/-10% DNI biases, thereby providing confidence to users of the reliability of the NSRDB (1998–2015) data set. Further, a comprehensive uncertainty estimate demonstrated the performance characteristics of the NSRDB (1998–2015) when including all sources of uncertainty at various time-averaged periods, a method that is not often used in model evaluations. Also, the study analyzed the inter-annual variability and mean-anomaly of NSRDB. When compared to ground-based measurements the inter-annual variability in the GHI and DNI for NSRDB (1998–2015) agreed to within 5%. This provides confidence that the NSRDB represents the surface measurements when annual estimates of solar radiation are considered. The study also found, the year 2015 appeared to have a 5%–10% reduction in solar irradiance (GHI) compared to other years (see e.g. fig. for Desert Rock, NV) for the central and southern plains. We will present details of the comprehensive NSRDB comparison and the results summarized above.