Title: Scenario Generation of Wind Farm Power for Real-Time System Operation

Abstract: This work proposes a method of wind farm scenario generation to support real-time optimization tools and presents key findings therein. This work draws upon work from the literature and presents an efficient and scalable method for producing an adequate number of scenarios for a large fleet of wind farms while capturing both spatial and temporal dependencies. The method makes probabilistic forecasts using conditional heteroscedastic regression for each wind farm and time horizon. Past training data is transformed (using the probabilistic forecasting models) into standard normal samples. A Gaussian copula is estimated from the normalized samples and used in real-time to enforce proper spatial and temporal dependencies. The method is evaluated using historical data from MISO and performance within the MISO real-time look-ahead framework is discussed.