Probabilistic determination of critical states in electrical grids using ensemble-based power forecasts

This work presents a new methodology for the detection of critical states in electrical grids by deriving probabilistic predictions from an ensemble of day-ahead weather forecast simulations. This multi-step approach will help transmission and distribution system operators in their decision making process. The method starts by using weather forecast simulation ensembles to derive power forecasts of the grid’s distributed energy resources, and running AC power flows to determine the state of the electrical grid for each ensemble member. This discrete collection of possible electrical grid states is converted into a probability distribution for key quantities of interest, using either kernel density estimation or by fitting to a candidate distribution model. A set of forecasting performance criteria is used to compare the kernel density approach and the fit models with one another. Finally, we demonstrate potential applications in a grid operation setting, including a concept for a forecasting application with a graphical user interface as well as an alert system for identifying potential grid constraint violations.