Graph decomposition for constructing blackstart restoration strategies in benchmark cases

Abstract

This paper describes methods for constructing benchmark cases and solution strategies related to restoring an electric grid model from a blackstart condition. Existing public cases for grid restoration problems are limited in size and scope—this paper delineates the features required to create new synthetic, detailed datasets for this purpose. As part of validating these datasets, benchmark results are included for an electric grid restoration strategy based on a formulation of the restoration problem as a time sequence of ac power flow solutions. The algorithm produces a benchmark restoration sequence for an example 200-bus case, showing how critical loads can be restored within 45 min, the majority of the grid within 6 h, and all loads within 36 h. At the core of the solution strategy is a directed graph decomposition heuristic, as the algorithm builds a bus energization spanning forest. Because these datasets are built synthetically, they can have a high degree of realism in features such as geographic coordinates and modeling complexities, while avoiding concerns of data confidentiality, meaning that the data produced in this work can be made publicly available for the benefit of the research community.