Deep Learning to Forecast Solar Irradiance Using a Six-Month UTSA SkyImager Dataset
| dc.contributor.author | Moncada, Ariana | |
| dc.contributor.author | Richardson, Walter | |
| dc.contributor.author | Vega-Avila, Rolando | |
| dc.date.accessioned | 2021-04-19T15:09:48Z | |
| dc.date.available | 2021-04-19T15:09:48Z | |
| dc.date.issued | 2018-07-31 | |
| dc.date.updated | 2021-04-19T15:09:48Z | |
| dc.description.abstract | Distributed PV power generation necessitates both intra-hour and day-ahead forecasting of solar irradiance. The UTSA SkyImager is an inexpensive all-sky imaging system built using a Raspberry Pi computer with camera. Reconfigurable for different operational environments, it has been deployed at the National Renewable Energy Laboratory (NREL), Joint Base San Antonio, and two locations in the Canary Islands. The original design used optical flow to extrapolate cloud positions, followed by ray-tracing to predict shadow locations on solar panels. The latter problem is mathematically ill-posed. This paper details an alternative strategy that uses artificial intelligence (AI) to forecast irradiance directly from an extracted subimage surrounding the sun. Several different AI models are compared including Deep Learning and Gradient Boosted Trees. Results and error metrics are presented for a total of 147 days of NREL data collected during the period from October 2015 to May 2016. | |
| dc.description.department | Mathematics | |
| dc.identifier | doi: 10.3390/en11081988 | |
| dc.identifier.citation | Energies 11 (8): 1988 (2018) | |
| dc.identifier.uri | https://hdl.handle.net/20.500.12588/418 | |
| dc.rights | Attribution 4.0 United States | |
| dc.rights.uri | https://creativecommons.org/licenses/by/4.0/ | |
| dc.subject | solar irradiance forecasting | |
| dc.subject | all-sky imaging | |
| dc.subject | optical flow | |
| dc.subject | artificial intelligence | |
| dc.subject | deep learning | |
| dc.subject | decision tree learning | |
| dc.title | Deep Learning to Forecast Solar Irradiance Using a Six-Month UTSA SkyImager Dataset | |
| dc.type | Article |