Predicting rare events of solar power production with the analog ensemble

Some revised versions of the analog ensemble (AnEn) algorithm to improve its performance when predicting rare events of solar power production are presented. Rare events are considered both extremely cloudy situations and clear sky conditions with very low aerosol concentrations. In these cases, the solar power production can be respectively unusually low or high, making it more difficult to find good analog dates in a historical archive of past forecasts. The lack of suitable quality analogs determines conditional biases from the AnEn when predicting these rare events. Here, an application of the AnEn to predict solar power generated by two photovoltaic plants located in Kuwait is presented. The modified versions of the AnEn are compared to the standard algorithm. We focus the verification of the AnEn's performance over the rare conditions outlining the improvements achieved with the new AnEn models.