Intra-hour cloud index forecasting with data assimilation

Abstract

We introduce a computational framework to forecast cloud index (CI) fields for up to one hour on a spatial domain that covers a city. Such intra-hour CI forecasts are important to produce solar power forecasts of utility scale solar power and distributed rooftop solar. Our method combines a 2D advection model with cloud motion vectors (CMVs) derived from a mesoscale numerical weather prediction (NWP) model and sparse optical flow acting on successive, geostationary satellite images. We use ensemble data assimilation to combine these sources of cloud motion information based on the uncertainty of each data source. Our technique produces forecasts that have similar or lower root mean square error than reference techniques that use only optical flow, NWP CMV fields, or persistence. We describe how the method operates on three representative case studies and present results from 39 cloudy days.fields derived hourly from a mesoscale NWP model. These two data sources are assimilated into a background ensemble that is initialized with a NWP 25 CMV field. We refer to the system as ANOC for the Assimilation of NWP winds and Optical flow CMVs.Generically, DA is a Bayesian technique to update numerical models using sparse and noisy obser-30 vations (Reich and Cotter, 2015;Asch et al., 2016). We use an ensemble Kalman filter (EnKF) (see, e.g., Evensen (2009)) to perform our assimilations. EnKFs are computational tools for DA where forecast uncertainty is represented by an ensemble. 35Optical flow is a method to determine a velocity field from consecutive scalar fields. Numerical methods for optical flow can be divided into two categories: dense optical flow (Horn and Schunck, 1981), where an entire vector field is produced, 40 and sparse optical flow (Lucas and Kanade, 1981), where point estimates of a vector field are produced. We use both dense and sparse optical flow to determine CMVs in this study.Advection of satellite-derived cloud properties for 45 intra-hour CI or irradiance field forecasts for solar power applications has been considered in several

The Seamless Solar Radiation (SESORA) Forecast for Solar Surface Irradiance—Method and Validation

“…Another great advantage of the optical flow method is that the algorithm of TV-L 1 (Method based on total variation in the regularization term and the L 1 -norm in the data fidelity term) is open source and a large community is constantly working on improving this and other methods by OpenCV (Open Source Computer Vision) [17]. There are many successful solar radiation forecasts that have been published in recent years that use cloud tracking methods with either geostationary satellites [3,[18][19][20], total sky imagery [21], or ground sensors [22]. One method of cloud tracking is to derive cloud motion vectors (CMV) from satellite imagery.…”