Framework for Training and Deployment Machine Learning Methods in Real-Time Simulator: Short-Term Kinetic Energy Forecasting in Power Systems
Jose Miguel Riquelme-Dominguez (University of Seville)
Francisco Gonzalez-Longatt (Loughborough University)
Jose M. Valles (Universidad Nacional Autónoma de México)
Jose Rueda Torres (TU Delft - Intelligent Electrical Power Grids)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Low-inertia power systems require more innovative operation, control, and protection strategies to maintain the operation secure and reliable. One of the challenges related to these systems is the need for knowledge of the inertia level (kinetic energy stored in the rotating masses) in real time. This paper proposes a framework for training and deploying machine learning methods for real-time power systems’ kinetic energy forecasting. Linear Regression and Long Short-Term Memory methods are implemented in the Python interpreter of the Typhoon HIL 404 real-time simulator for forecasting the kinetic energy of the Nordic Power System in real-time. This paper provides implementation details together with possible future expansions of the framework. Simulation results show that the trained models can predict the kinetic energy in a forecasting horizon of four hours with a Mean Absolute Error lower than other methods currently available in the literature.