Forecasting Variability in Local Area Photovoltaic Power Generation using the Dynamic Mode Decomposition

We propose and investigate the application of the Dynamic Mode Decomposition to the characterization of the dynamics of fair-weather convective Cumulus clouds. The shadows of such clouds cause variability in the electric power output of solar photo-voltaic facilities; consequently, the ability to predict the short-term dynamics of such clouds in addition to their simple advection by wind may be useful in mitigating variability of solar power generators. In this study, cloud dynamics are captured by sequences of visible-light photographic video frames. The most obvious challenge to this technique is that spatiotemporal dynamics of the evolution of a cloud may not be sufficiently spanned by images collected from just one camera, in which case multiple perspectives or a higher state dimension may be needed.