GOES-based solar energy prediction products for decision makers

Renewable energy capabilities such as solar and wind have emerged over recent years as a national priority. Owing to the inherent volatility and of these renewable energy resources with respect to conventional fossil-fuel resources, facilities face considerable challenges in managing their power production. These operational user requirements translate to outstanding research needs for improved predictive capabilities from both models and observations.

Solar irradiance at the surface varies dramatically in the presence of cloud cover. Work has begun to develop a short-term predictive capability for cloud cover impacts on the availability of the solar resource (direct and diffuse components). This research focuses on GOES-derived cloud products for i) short-term (< 3 hr) predictions of solar irradiance variability at the surface, and ii) the statistical likelihood of cloud cover at mid- to long-term (> 3 hrs to several days) time scales based on satellite cloud cover statistics conditions on meteorological flow regimes. For short-term predictions, we use GOES imagery to identify cloud locations and optical/geometric properties. A cloud advection method is applied to predict short-term cloud movement under simplifying assumptions. Downwelling surface irradiance, based on predicted cloud locations and properties (e.g., optical thickness, particle size, and cloud top height), is then computed for these advected fields to yield a time series at selected sites. For mid- to long-term predictions, we leverage a regional satellite cloud climatology dataset (compiled over 12 years), conditioned by numerical weather prediction model-predicted wind flow regimes, to estimate cloud cover probability and related statistics. Results are available on hourly intervals over the duration of the forecast period.

Both thrusts of this research provide novel, quantitative utilities that speak to important and outstanding predictive needs of this nascent industry.