An Optimization-Based DC-Network Reduction Method

Reduced order models of dc networks using Ward, REI, and bus-aggregation approaches have been used to advantage for some time. Proposed in this paper is an optimization-based framework that can easily be modified to generate a Ward reduction, a bus-aggregation-based reduction, or a range of hybrid reductions between these two extremes, generating reduced-order models that are tailored to the purpose of the study. The so-called optimization-based Ward reduction method is demonstrated to be numerically stable on practical examples and is specifically applied to the problem of large-reactance branch elimination. We show that it offers accuracy advantages over the traditional approach of removing high-reactance branches using a threshold criterion.