Sarvarizadeh Kouhpaye, Miad Rajabdorri, Mohammad Lobato Miguélez, Enrique Sigrist, Lukas Kazemtabrizi, Behzad Troffaes, Matthias 2025-09-15T07:56:54Z 2025-09-15T07:56:54Z 2025-09-01 0142-0615 https:doi.org10.1016j.ijepes.2025.110884 http://hdl.handle.net/11531/104025 Artículos en revistas The transition from fossil fuels to renewable energy sources in power systems has resulted in lower system inertia and deteriorated frequency response characteristics. The challenge becomes even more pronounced in island power systems, where low inertia and limited frequency control capacity are already issues. Despite the existing preventive measures to maintain frequency limits, under-frequency load-shedding (UFLS) remains a crucial corrective action. In small isolated power systems, even with high amounts of reserve, big outages will lead to inevitable UFLS due to scarcity of inertia and primary frequency control capacity. Using a data-driven regression tree method, this paper introduces a novel UFLS-aware unit commitment formulation that can accurately estimate the amount of resulting UFLS for every possible outage. The estimations are then utilized to relax the reserve requirement constraint and lower the operation costs. The proposed formulation enables co-optimizing operation costs and UFLS to minimize the total cost. Additionally, the response speed of generating units is considered, ensuring timely reserve delivery. The efficacy of the proposed method is demonstrated through simulations on the model of a Spanish island power system, highlighting potential reductions in operation costs and system security improvements. en-GB Revista: International Journal of Electrical Power & Energy Systems, Periodo: 1, Volumen: online, Número: , Página inicial: 110884-1, Página final: 110884-15 Instituto de Investigación Tecnológica (IIT) Under frequency load shedding aware unit commitment in island power systems info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion info:eu-repo/semantics/openAccess Under-frequency load-shedding; Unit commitmentIsland power systems; Machine learning; Logistic regressionRegression tree