An Operational Forecasting System for Renewable Power in the Southwest US Using NWP, Satellite Imagery, and DG PV Production Data

We developed an operational renewable power forecasting system for utilities in the Southwestern US using a combination of high-resolution numerical weather prediction, satellite imagery, distributed generation solar power production data, and irradiance sensors. The system generates forecasts with 10 second resolution for the first 30 minutes and 3 minute resolution out to 3 days. First, we use the WRF model to produce multi-day forecasts of temperature, wind speed, and irradiance. We then use real-time solar and wind power production data and the plant specifications to convert the model data into power predictions. The model uses a high spatial resolution (1.8km) configuration that performs well in the arid, complex terrain of the Southwestern US. We use 3 minute output from the WRF model so that we can predict periods of high irradiance and solar power variability. Next, we implemented an irradiance forecast based on GOES imagery for 30 to 120 minute time horizons. Finally, we use a network of distributed generation photovoltaic installations and irradiance sensors to generate irradiance forecasts out to 30 minutes that are updated every 60 seconds. These short-term forecasts can help anticipate destabilizing solar ramp events, enable preemptive curtailment to avoid high ramp rates, and reduce the battery size needed to control ramp rates. We developed our forecasts in collaboration with local utilities Tucson Electric Power and Arizona Public Service.