I International Conference on Renewable Energy - ICEREH 2026, Valverde (España). 10-13 noviembre 2026
Resumen:
Small island power systems are vulnerable to extreme weather events that can split the system into asynchronous areas and cause severe frequency disturbances. Existing Frequency-Constrained Unit Commitment (FCUC) formulations secure the schedule only against outages within a single synchronous system, and therefore cannot guarantee frequency security once separation occurs. This paper proposes an areabased FCUC framework that secures the day-ahead schedule against separation-driven under-frequency events. Neural network classifiers are trained to capture the nonlinear postseparation frequency nadir of areas, while importance sampling is used to increase their accuracy near the security boundary. The framework is compared against standard FCUC on a model of the La Palma island grid. Results show that the proposed formulation ensures dynamic frequency security during separation events.
Resumen divulgativo:
Este artículo presenta un nuevo método de programación diaria para ayudar a las redes eléctricas insulares a mantener la estabilidad de frecuencia cuando fenómenos extremos separan la red en áreas. Las redes neuronales permiten mejorar la resiliencia y reducir el riesgo de cortes.
Palabras clave: resilience, unit commitment, frequency, machine learning, neural network
Fecha de publicación: 10-nov-2026
Cita:
M. Sarvarizadeh, B.V. Venkatasubramanian, M. Rajabdorri, L. Sigrist, E. Lobato, M. Panteli, "Resilient Day-ahead Scheduling of Island Grids Against Separation-Driven Under-Frequency Events", presentado en I International Conference on Renewable Energy - ICEREH 2026, Valverde, España, 10-13 noviembre 2026
IIT-26-285C
