Keywords
Instituto de Investigación Tecnológica (IIT) - Innovación docente y Analytics (GIIDA), adaptive network constraints, complexity reduc tion, optimization models, power system
Adaptive Complexity Reduction of Network Constraints for Large-Scale Power System Optimization Models
Abstract
Power system optimization models can quickly become computationally intractable, especially when spanning large geographic areas. Common approaches address this by uniformly simplifying the entire network, for example, reducing the DC Optimal Power Flow to a Transport Problem or a Single Node representation. In contrast, this paper introduces a novel framework for adaptive complexity reduction that assigns the level of spatial and technical detail on a line-by-line basis. The set of lines is partitioned into three subsets, each modeled with a different power flow representation, and the mathematical formulation is adjusted accordingly. Two exemplary assignment strategies are presented: a buffer-based approach with a Zone-Of-Interest, and a utilization-based method. Case studies on the extended NREL-118 bus test system using the open-source LEGO model, where we solve a generation expansion problem, show that an 81% reduction in computational effort can be achieved with only 3% regret relative to the full DC-OPF benchmark, which can be tightened to 1% while still achieving a 62% reduction. These results confirm that the framework can reduce computational effort significantly without degrading solution quality.
Power system optimization models can quickly become computationally intractable, especially when spanning large geographic areas. Common approaches address this by uniformly simplifying the entire network, for example, reducing the DC Optimal Power Flow to a Transport Problem or a Single Node representation. In contrast, this paper introduces a novel framework for adaptive complexity reduction that assigns the level of spatial and technical detail on a line-by-line basis. The set of lines is partitioned into three subsets, each modeled with a different power flow representation, and the mathematical formulation is adjusted accordingly. Two exemplary assignment strategies are presented: a buffer-based approach with a Zone-Of-Interest, and a utilization-based method. Case studies on the extended NREL-118 bus test system using the open-source LEGO model, where we solve a generation expansion problem, show that an 81% reduction in computational effort can be achieved with only 3% regret relative to the full DC-OPF benchmark, which can be tightened to 1% while still achieving a 62% reduction. These results confirm that the framework can reduce computational effort significantly without degrading solution quality.
Power system optimization models can quickly become computationally intractable, especially when spanning large geographic areas. Common approaches address this by uniformly simplifying the entire network, for example, reducing the DC Optimal Power Flow to a Transport Problem or a Single Node representation. In contrast, this paper introduces a novel framework for adaptive complexity reduction that assigns the level of spatial and technical detail on a line-by-line basis. The set of lines is partitioned into three subsets, each modeled with a different power flow representation, and the mathematical formulation is adjusted accordingly. Two exemplary assignment strategies are presented: a buffer-based approach with a Zone-Of-Interest, and a utilization-based method. Case studies on the extended NREL-118 bus test system using the open-source LEGO model, where we solve a generation expansion problem, show that an 81% reduction in computational effort can be achieved with only 3% regret relative to the full DC-OPF benchmark, which can be tightened to 1% while still achieving a 62% reduction. These results confirm that the framework can reduce computational effort significantly without degrading solution quality.
Información detallada
Materias, derechos, colecciones e identificadores