Hybrid solar forecasting method uses satellite imaging and ground telemetry as inputs to ANNs

Abstract

This work describes a new hybrid method that combines information from processed satellite images with Artificial Neural Networks (ANNs) for predicting global horizontal irradiance (GHI) at temporal horizons of 30, 60, 90, and 120min. The forecast model is applied to GHI data gathered from two distinct locations (Davis and Merced) that represent well the geographical distribution of solar irradiance in the San Joaquin Valley. The forecasting approach uses information gathered from satellite image analysis including velocimetry and cloud indexing as inputs to the ANN models. To the knowledge of the authors, this is the first attempt to hybridize stochastic learning and image processing approaches for solar irradiance forecasting. We compare the hybrid approaches using standard error metrics to quantify the forecasting skill for the several time horizons considered.