A0229
Title: Statistics of the power grid frequency
Authors: Dirk Witthaut - Forschungszentrum Juelich GmbH (Germany) [presenting]
Abstract: A reliable supply of electric power is vital for our society. The stable operation of the power system requires that generation and load are balanced, which is ensured by elaborate control systems. Temporary imbalances occur, both due to rapid stochastic fluctuations and slower effects of scheduling and electricity trading. Such imbalances manifest in the grid frequency, which is constantly monitored and used to control generation to restore the balance. We discuss the intricate statistical properties of frequency recordings from various grids around the world with a focus on the emergence of heavy tails and long correlations. We introduce a physics-inspired machine learning model that bridges time scales from seconds to hours: Stochastic differential equations describe the fast dynamics of the frequency and the control system, while artificial neural networks are used to incorporate external influences such as scheduling and trading. The model successfully reproduces important statistical properties of the frequency in the European grid and enables probabilistic forecasting.
