Operational Solar Forecasting System for DNI and GHI for Horizons Covering 5 Minutes to 72 Hours

Abstract:

I will describe the methodology used to develop and deploy operationally a comprehensive solar forecasting system for both concentrated

and non-concentrated solar technologies. This operational forecasting system ingests data from local telemetry, remote sensing and Numerical

Weather Prediction (NWP) models, processes all the diferent types of data (time series, sky images, satellite images, gridded data, etc.) to produce

concatenated solar forecasts from 5 minutes out to 72 hours into the future. Each forecast is optimized with stochastic learning techniques that include

input selection, model topology optimization, model output statistics, metric fitness optimization and machine learning.

These forecasts are used by solar generators (plant managers), utilities and independent system operators for operations, scheduling,

dispatching and market participation.