Probabilistic Real-time Price Forecast and the Application to Pumped Storage Hydro Unit Optimization

In electric power markets, a robust and reliable price forecasting model is a challenge and of interest for both power producers and consumers. In this paper, first, an ARIMAX based single point forecast methodology is developed. To reflect the uncertainty of the single point forecast, the predicted price variable in k-Iook ahead hour is transformed to a vector with Gaussian distribution. All the interdependence of forecasting error is saved in the form of a unique covariance matrix. Finally, by using a multivariate Gaussian random number generator, multiple statistical scenarios are generated to reflect the uncertainty associated with the single point price forecast. The developed probabilistic price forecasts and the single point price forecast are used in a Pumped Storage Hydro Unit (PSHU) profit maximization formulation and tested with real-time (RT) market rolling PSHU simulations. The results show that the probabilistic forecast can incorporate the RT price uncertainty and the stochastic profit maximization model with probabilistic pricing scenarios can consistently generate better RT profits for PSHU.