Evaluation of hours - ahead solar forecasting using satellite imagery and 1 Numerical Weather Prediction

Executive Summary 5 A method for solar forecasting using cloud motion vectors (CMV) from satellite imagery with the 6 ability to characterize forecast uncertainty has been developed. On average, the root mean square error 7 (RMSE) for CMV forecast increases with increasing forecast horizon and becomes larger than the North 8 American Model (NAM, a numerical weather prediction model) forecast error at between 6 hours and 1 9 day. Consequently, satellite CMV forecasts are superior for short time horizons and they are currently 10 used as the model of choice in SolarAnywhere forecasts up to a 6 hour horizon. However, the forecast 11 horizon at which the ‘crossover’ between CMV and NAM occurs is dynamic (as short as 2 hours) and 12 could be adjusted if the CMV forecast certainty was known at forecast issue time. The RMSE of CMV 13 forecasts was most related to satellite image entropy and uniformity. The relative performance (RP), i.e. 14 the ratio of the errors of NAM and satellite CMV forecast is also analyzed. The average rRMSE of 15 predicting RP is shown to be about 30% for two different modeling techniques. The models and metrics 16 developed in this project can be applied to choose the optimal forecast model and reduce solar forecast 17 errors especially for hour-ahead forecasts. An operational CMV forecast can be produced with a 25 18 minute latency. 19