We study a number of methods to produce and improve renewable power forecasts at many time scales. Below are descriptions of some of the methods we use. More information on these topics can be found in our presentations and papers.
The University of Arizona Department of Hydrology & Atmospheric Sciences runs the Weather Research and Forecasting (WRF) model with a configuration customized for the Southwest United States. The model runs in a nested configuration with a 5.4 km outer domain grid spacing that covers the Western US and a 1.8 km inner domain that covers Arizona and New Mexico. An ensemble of runs are produced each day based on different initializations. These forecasts are best suited for forecasting events 6 hours to 8 days in the future. Graphics from the most recent WRF runs can be found here.
Satellite-derived Irradiance Forecasts
For one hour to four hour forecast horizons, we rely on irradiance estimates and forecasts derived from geostationary satellite images. We currently use a physical model that determines the cloud properties from the visible and infrared images from the GOES-W satellite. We can then use these cloud properties to estimate the irradiance to to produce a forecast.
Sensor Network Forecasts
Using a network of irradiance sensors and rooftop PV power data that acts as a proxy for irradiance, we produce short-term, irradiance forecasts for the Tucson area. We've shown that these forecasts have skill over persistence forecasts in the one minute to thirty minute forecast horizon zone.
Lorenzo et. al. Solar Energy 122 1158-1169, Irradiance Forecasts based on an Irradiance Monitoring Network, Cloud Motion, and Spatial Averaging (2015). journal link
Satellite Irradiance Optimal Interpolation
We've also used the network of sensors in Tucson to improve satellite derived irradiance estimates via a method know as optimal interpolation. This method eliminates bias in the estimates and reduces the root mean squared error of the estimates by 50%.