Analytical Threat Modeling for Power Systems

The vulnerability analysis of electrical power systems has become a growing field of research in past years. The transformation of power systems into Cyber Physical Energy Systems and geopolitical developments stress the importance of analyzing possible weaknesses of power systems as a part of critical infrastructure. In this paper, a two-step bilevel network interdiction modeling approach is developed and tested. It consists of a bilevel optimization model to identify high-impact-low-probability scenarios (Filtering model) and a single-level optimization model to assess the impact of the identified scenarios (Ranking model). The Filtering model is setup such that it can easily be extended by implementing additional features of a network interdiction model as services. In case studies for the validation of the two presented models, three test grids are considered. All case studies are based on a Python implementation of the models to be applicable to pandapower grids. The impact of the solver choice is extensively studied, stressing the importance of the applied combination of bilevel and lower-level solver. Relevant criteria for solver combinations to prevent incorrect results are identified. Applying the appropriate solvers to the services used, the Filtering model and its single-level reformulation serve as a bilevel network interdiction model for pandapower grids. The extension of the model including additional services is examined and the importance of the Ranking model for quantifying the impact of defined scenarios is shown which is consistent with the literature.