Regime-Dependent Short-Range Solar Irradiance Forecasting

• DOI: 10.1175/jamc-d-15-0354.1
• Publication Date: 2016-04-12T22:36:29Z
• Authors: T. C. McCandless; G. S. Young; S. E. Haupt; L. M. Hinkelman
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Abstract This paper describes the development and testing of a cloud-regime-dependent short-range solar irradiance forecasting system for predictions 15-min-average clearness index (global horizontal irradiance). regime-dependent artificial neural network (RD-ANN) classifies cloud regimes with k -means algorithm on basis combination surface weather observations, GOES-East satellite data. The ANNs are then trained each regime to predict index. RD-ANN improves over mean absolute error baseline clearness-index persistence by 1.0%, 21.0%, 26.4%, 27.4% at 15-, 60-, 120-, 180-min forecast lead times, respectively. In addition, version this method configured variability predicts more accurately than does smart technique.