Optimizing Feeder Selection Across Underfrequency Load Shedding Scheme Steps

Abstract

This paper addresses the problem of optimal feeder selection in underfrequency load shedding (UFLS) schemes. Conventional UFLS schemes, which remain the predominant approach in power systems, operate in predefined steps, with each step intended to disconnect a specific percentage of the total load. However, the increasing variability in load profiles, driven largely by the growing penetration of distributed generation, makesit increasingly challenging to consistently achieve the targeted shedding levels. To overcome these challenges, we formulate the feeder selection task as a mixed-integer linear programming (MILP) problem, enabling the optimal assignment of feeders to each UFLS stage. Given the computational burden of the full MILP model, especially in large-scale systems with multiple stages and many feeders, we propose a sequential optimizationapproach that solves each UFLS step independently. Both models are tested using real historical hourly data from a region in Spain. Results show that the sequential model achieves nearly the same solution quality as the full MILP formulation, with significantly lower computational effort, making it suitable for practical deployment in operational settings.

This paper addresses the problem of optimal feeder selection in underfrequency load shedding (UFLS) schemes. Conventional UFLS schemes, which remain the predominant approach in power systems, operate in predefined steps, with each step intended to disconnect a specific percentage of the total load. However, the increasing variability in load profiles, driven largely by the growing penetration of distributed generation, makesit increasingly challenging to consistently achieve the targeted shedding levels. To overcome these challenges, we formulate the feeder selection task as a mixed-integer linear programming (MILP) problem, enabling the optimal assignment of feeders to each UFLS stage. Given the computational burden of the full MILP model, especially in large-scale systems with multiple stages and many feeders, we propose a sequential optimizationapproach that solves each UFLS step independently. Both models are tested using real historical hourly data from a region in Spain. Results show that the sequential model achieves nearly the same solution quality as the full MILP formulation, with significantly lower computational effort, making it suitable for practical deployment in operational settings.