Abstract: Utility-Scale Intra-Hour Forecasting Using Two Sky Imagers (92nd American Meteorological Society Annual Meeting (January 22-26, 2012))

Wednesday, 25 January 2012: 2:00 PM

Utility-Scale Intra-Hour Forecasting Using Two Sky Imagers

Room 345 (New Orleans Convention Center )

Bryan Urquhart, Univ. of California, La Jolla, CA; and C. W. Chow, J. Kleissl, J. L. Bosch, D. Jeon, and J. Blatchford

Recorded Presentation

A pair of sky imagers is being deployed at a 48MW photovoltaic (PV) facility in Henderson, Nevada. The sky images will be analyzed using image processing techniques developed by Chow et al. (in review) for short term forecasting of global horizontal irradiance (GHI). Cloud decision images generated from the two instruments (Fig. 1c) will be combined to generate a single large cloud map over the site to provide full coverage of the 1.7 km² plant (Fig. 2). The cloud map concept is shown in Fig. 3 for the single sky imager deployed at the University of California, San Diego.

Figure 1 A raw sky image (a) with corresponding red-channel to blue-channel ratio (b) is converted to a cloud decision image (c) for cloud mapping. Figure from Urquhart et al. (2011).

Figure 2 Deployment configuration (red dots) for imagers at the 48 MW PV plant. Image from Google Maps.

Figure 3 Projected cloud map over the University of California, San Diego campus. Image generated in Google Earth.

Using the cloud maps, along with topography and solar geometry, shadow maps at a granular resolution of 100 m² will be constructed over the facility to assess the capability of sky imagery to provide high spatial and temporal resolution irradiance estimates for a large scale PV farm. DC power data from 110 inverters, 5 plane-of-array silicon pyranometers, and 15 calibrated reference PV cells will be used to map power output and irradiance to validate the ground irradiance estimates from the sky imager pair.

The short term ramp forecasting from 30 seconds to 20 minutes ahead over the entire site can improve grid integration of the variable solar resource by increasing situational awareness of plant operators ahead of rapid production changes due to clouds. Further value will be added with the FERC NOPR of 15 minute transmission scheduling intervals. The ability to leverage actual inverter-level power data from a large scale PV plant provides this study a one-of-a-kind opportunity to test short term forecasting algorithms in a real utility scale environment.

Supplementary URL:


Figure 1 A raw sky image (a) with corresponding red-channel to blue-channel ratio (b) is converted to a cloud decision image (c) for cloud mapping. Figure from Urquhart et al. (2011).