Research Scientist National Renewable Energy Laboratory Golden, Colorado, United States
For over 20 years, the National Solar Radiation Database (NSRDB), covering most of the western hemisphere, has been a source of public data for many solar energy applications. Recent improvements in satellite technology and machine-learning-based remote sensing methods have added tremendous value to the NSRDB in terms of both the quantity and quality of the data.
For example, the historical NSRDB data that is available from 1998 to present with one year lag is processed on a nominal 4x4 km grid spacing at a 30min frequency. Beginning in 2018, the NSRDB has additional datasets at 2x2 km 5min resolution available for the Continental United States, Hawaii, Mexico, and the Caribbean Islands, and at a 2x2 km 10min resolution available for North and South America from +60 to -60 degrees latitude. The improved spatiotemporal resolution should be a great asset to our stakeholders, especially for the analysis of utility scale solar installations which typically desire a higher resolution than the previously available 4x4 km 30min data.
Moreover, we have developed new methods for the prediction of cloud properties from satellite data using physics-guided machine learning. These methods were originally developed to compensate for the limitations of traditional cloud property retrieval algorithms, but they have proven to be generally more accurate than the traditional algorithms. The results demonstrate higher accuracy in the modeled irradiance that is expected to be helpful for a wide variety of solar energy applications.
In summary, the goal of the NSRDB is to provide the public with the highest-quality freely-available solar irradiance data possible. In this context, the NSRDB continues to evolve and push the envelope of what a public solar dataset can be. We think these recent advancements are important contributions to the solar energy community, and we hope that they will be fully taken advantage of.
