EVALUATION OF NUMERICAL WEATHER PREDICTION FOR SOLAR FORECASTING

Numerical weather prediction models are generally the most accurate tools for forecasting solar irradiation several hours in advance. This study validated the North American Model, Global Forecast System, and European Centre for Medium-Range Weather Forecasts global horizontal irradiance forecasts against ground measurement data and persistence forecasts. A ll numerical weather prediction models were biased for measured clear conditions by less than 50 watts per metre and for cloudy conditions the models were biased towards forecasting clear conditions, resulting in large, positive biases. Mean bias errors were obtained for each numerical weather prediction model as a function of solar zenith angle and forecast clear sky index. The North American Model and Global Forecast System were positively biased by up to 150 watts per metre for forecast clear sky conditions, while the European Centre for Medium-Range Weather Forecast mean bias errors were small. Outside of the few clear forecasts that were actually cloudy, the reason for this bias was that the Global Forecasting System and especially the North American Model forecasts can exceed clear sky irradiances by up to 40 percent, indicating an inaccurate clear sky model. For forecast cloudy conditions the North American Model and Global Forecast System models had a moderate negative bias, while European Centre for Medium-Range Weather Forecast forecasts were most biased. Model output statistics-corrected numerical weather prediction forecasts provided important baseline accuracy for evaluating other forecasting techniques. Model output statistics minimized mean bias errors for all numerical weather prediction models. Root mean square errors were also reduced, especially for intermediate clear sky indices. The model output statistics-corrected Global Forecast System provided the best solar forecasts for the continental United States. The European Centre for Medium-Range Weather Forecast provided the most accurate forecast in cloudy conditions, while the Global Forecast System had the best clear sky accuracy. Please use the following citation for this report: Mathiesen, Patrick, Jan Kleissl. 2012. Evaluation of Numerical Weather Prediction for Solar Forecasting. California Energy Commission. Publication number: CEC-500-2013-115.