A decision support framework for node vulnerability analysis using multi-view similarity fusion: application to the Venezuelan power grid

Node vulnerability in electric power grids arises from multiple mechanisms, including network topology, power-flow redistribution, and cascading failures. Because these mechanisms provide complementary information, no single vulnerability metric offers a complete basis for operational decisions. This paper proposes a similarity-based, multi-view decision support framework that integrates topological and load-flow-derived representations using Similarity Network Fusion (SNF) to produce consensus-based node criticality rankings. Rather than combining independent scalar indices, the proposed approach fuses relational similarity patterns across heterogeneous data views, enabling non-compensatory, data-driven prioritization of infrastructure interventions. The methodology is evaluated on the Venezuelan 80-node, 107-line transmission network using cascading-failure simulations incorporating islanding and overload-based line tripping. Damage is quantified using the normalized Area Under the Damage Curve (AUC) and the final fraction of unserved load. Results show that vulnerability is governed primarily by power-flow-related effects rather than topology alone. At the maximum tested contingency (10 removed nodes), the fusion of DC power-flow-based views leaves 77% of load unserved, compared with 28% for topology-based views (normalized AUC: 1.00 vs 0.22). Integrating both perspectives yields a consensus ranking (37% unserved load; normalized AUC 0.54), demonstrating the value of multi-view data fusion for infrastructure decision support.