Identifying Vulnerabilities in the Dutch Electricity Grid

Abstract

Electricity networks are becoming increasingly critical to society, while recent sabotage incidents demonstrate their exposure to deliberate physical attacks. Distribution System Operators (DSOs) therefore need to understand both which assets are most vulnerable to malicious threats and which are structurally most critical to network functioning. Existing research predominantly approaches this problem from a structural perspective using graph-theoretic methods, while asset-specific vulnerability to malicious physical threats has received considerably less attention. Consequently, it remains unclear whether structurally critical assets are also the assets most vulnerable to deliberate physical attacks. This research addresses this gap through a case study of Stedin’s subtransmission network. Two complementary assessment frameworks were developed. First, an asset-specific Vulnerability Index (VI) was constructed based on characteristics relevant to deliberate physical threats, identified through literature and validated with Stedin experts. The VI combines asset exposure with the potential impact of asset failure. Second, a structural criticality framework was developed using graph-theoretic centrality measures, N − 1 contingency analysis and sequential targeted attack simulations. Applying both frameworks to the same network enabled direct comparison between asset-specific vulnerability and structural criticality.

The results demonstrate that vulnerability to malicious physical threats cannot be explained by network topology alone. Relevant asset characteristics were identified across four dimensions: Topological, Spatial/Physical, Operational and Security. These characteristics do not contribute equally to vulnerability. Physical redundancy and N − 1 capacity redundancy emerged as the strongest and most consistent determinants, demonstrating that the quality of redundancy is more important than its mere topological presence. In particular, an alternative connection may exist structurally without being physically independent or having sufficient capacity to supply the affected demand after an outage.

The structural assessment shows that criticality in a predominantly radial distribution network cannot be captured by a single decision rule alone. The N − 1 contingency analysis identifies individual assets whose removal causes islanding and loss of weighted connectivity, while the targeted attack simulations show how network performance deteriorates under sequential removals. Nodes generally represent more effective targets than individual connections, as their removal can disconnect multiple downstream branches simultaneously.

The targeted attack simulations further show that the most effective decision rule depends on the attacker’s objective and available information. Impact-informed IEMDE causes the largest initial reduction in remaining weighted connectivity for both nodes and edges. However, additional impact information does not necessarily result in faster overall fragmentation. For nodes, IEMDE