Objective & Background
The variability of wind and solar power generation is often mitigated when solar, and power plants are installed in nearby locations and are paired with energy accumulation systems. For example, in countries such as Kuwait having facilities like Shagaya with both PV panels and wind turbines allows having a continuous generation of renewable energy throughout the day. Especially in desert areas, winds associated with low-level jets permit the wind power production during night-time when the solar generation is missing. The National Center for Atmospheric Research (NCAR) has developed a system to generate wind and probabilistic solar predictions for the Shagaya facility located in a desert area in Kuwait. These predictions are based on the analog ensemble (AnEn) technique [1] which post-processes wind and solar irradiance predictions based on the Weather and Research Forecasting System (WRF) numerical model. The ensemble forecasts are made of twenty members and are generated independently for the wind and solar power. We hereby present a method to pair the ensemble members from the two independent systems to obtain a unique ensemble prediction of the aggregated wind and solar generation.
Method
The ensemble members provided by the AnEn are statistically indistinguishable, and they are generated without space-time correlation. We apply the Schaake Shuffle (SS) technique [2], widely used for hydrological application, to reorder the ensemble members and recover space-time variability of solar and wind power forecast time-series. In this technique, the ensemble members for a given forecast lead-time are ranked and matched with the rank of solar or wind power past observations at the same hours appropriately selected across the historical record. The ensembles are then reordered to match the original order of the selected historical data. Using this technique, the observed inter-plants correlation and the observed temporal autocorrelation are almost entirely recovered. After the reordering through the SS, the paired solar and wind power members can be summed to build a unique ensemble of combined generation.
Principal Findings
The combined solar/wind power ensemble prediction obtained by using the AnEn and the SS techniques shows an excellent statistical consistency verified by compiling rank histograms, spread/skill, and reliability diagrams.
Conclusion
Reliable quantification of the uncertainty of the total power production of the Shagaya plant is provided by the ensemble forecasting system obtained by the combination of the AnEn and SS technique. Also, each ensemble member preserves the temporal autocorrelation of the observations. Hence, the ensemble forecast can be used to optimize the charging/discharging cycles when an accumulation system is also available.