A Stochastic Approach for Downscaling Solar Irradiance Data Products

• Activity Number: 395 - Statistical Models for High-Dimensional Computer Output
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Wenqi Zhang* and William Kleiber
• Companies: University of Colorado at Boulder and University of Colorado
• Keywords: Downscaling; Solar Irradiance; Solar Power; Renewable Energy; High-dimensional computer output

Abstract

Solar power as a source of electricity is increasingly viable, and solar photovoltaic (PV) installations are becoming more common. However, PV planning studies require high frequency solar irradiance scenarios to understand potential variability and intermittency. Existing remote sensing solar data products are often available over large spatial domains, but are limited in their temporal resolution. For instance, the global horizontal component (GHI) in National Solar Radiation Database (NSRDB) is available at time resolution of 30 min on a 4 km grid. We develop an algorithm to stochastically downscale GHI from the NSRDB to the 1 minute resolution. We illustrate the performance of the algorithm at a set of sample locations around Oregon, US. Simulated ensembles show good coverage properties, and maintain important temporal correlation structures. The resulting downscaled ensembles allow for understanding the nonlinear variability inherent in GHI at locations without direct measurements.