Benchmarks

Power-grid cascade containment with named public benchmarks.

BlueChips evaluated N-1-1 contingency analysis across IEEE 140-bus and PEGASE 1,354-bus networks. The benchmark asks whether a method can certify cascade containment before failure propagates.

Headline result

On the PEGASE 1,354-bus network, BlueChips contained 79.1% of cascades.

Graph Lasso dropped to 0.8% on the same scale regime, and a deep learning GCN did not scale to the 1,354-bus benchmark. The result is stated this way so the benchmark can be quoted without losing the network name, method class, or comparison point.

BlueChips FIBI-Penalised

79.1% PEGASE containment 1,354-bus network

Graph Lasso

0.8% PEGASE containment Degraded at scale

Deep Learning GCN

N/A PEGASE containment Failed to scale
Method Cascade Success Rate Computation Time Scales to 1,354-Bus?
BlueChips FIBI-Penalised 79.1% 2.5ms Yes
BlueChips FIBI-Localised 67.3% 2.5ms Yes
Graph TV 80.1% N/A Yes
Spectral Clustering 33.2% 21ms Partial
Deep Learning GCN 14.5% 101ms No
Graph Lasso 0.8% N/A Yes, degraded

Benchmark scope: IEEE 140-bus and PEGASE 1,354-bus networks; 10,612 and 4,044 N-1-1 contingency instances respectively; 14,656 total fault scenarios. CSR = Cascade Success Rate. Containment = fraction of fault impact isolated.