Finding Critical Nodes in Interdependent Networks with SAT and ILP Solvers

TL;DR

This paper presents a novel method combining SAT and ILP solvers to identify critical nodes in interdependent networks, effectively modeling cascading failures and solving an NP-hard problem for moderate-sized systems.

Contribution

It introduces a two-phase approach using SAT and ILP to accurately find critical nodes in interdependent networks, improving feasibility for moderate problem sizes.

Findings

Feasible for networks with up to 30 nodes

Effective in modeling cascading failures

Provides a baseline for heuristic comparison

Abstract

Infrastructure systems, such as power systems, often experience cascading failures. Modeling an infrastructure system as a collection of interdependent networks has recently received attention as a way to explain cascading failures. In this study, we propose an approach to find the set of critical nodes in an interdependent network. For an integer k, we say that a set of k nodes is critical if the initial failures of these k nodes result in the most severe cascading failure among all sets of k nodes. This approach adopts the seminal model of interdependent networks proposed by Buldyrev et al., in which new link failures occur in a network if the connectivity is lost in the paired network. The problem of finding critical nodes is NP-hard; thus the aim of the approach is to accurately solve the problem in feasible time for moderate-size problem instances. The proposed approach consists of two phases. In the first phase, the maximum number of failure propagation stages is computed by repeatedly solving the Boolean satisfiability problem. This number is then used in the second phase, where the set of critical nodes is computed using integer linear programming. The results of applying this approach to a variety of problem instances demonstrate that the approach is feasible for up to at least 30 nodes and can be used as the baseline to compare the performance of heuristic solutions.