Gallery
Description
⚡ Resilien-T GridProject OverviewHigh-voltage electrical transmission grids are increasingly exposed to extreme, climate-driven hydrogeological threats. Traditional AI monitoring often acts as a "black box" and completely ignores both the physical laws of soil mechanics and the underlying topological structure of the network. Resilien-T Grid solves this by implementing a Physics-Informed Graph Convolutional Network (GCN) architecture. It serves as an advanced Digital Twin that ingests real-time weather evolution data and satellite radar interferometry (InSAR ground displacement) to compute a structural Systemic Risk Index (SRI) for every grid component, automatically triggering network flow countermeasures to bypass compromised bottlenecks and prevent regional cascading blackouts.Core Features & Technical InnovationBuilt completely from scratch using pure linear algebra in NumPy, the core engine performs spatial message-passing directly over the regional power grid topology, avoiding neural network hallucinations by computing a custom symmetric normalized matrix to aggregate geo-climatic risk features from neighboring connected towers. The GCN hazard state is tightly coupled with deterministic soil mechanics using the Infinite Slope Stability Model, where pore water pressure dynamically increases based on cumulative rainfall load while soil cohesion decays exponentially as a function of millimetric satellite-recorded InSAR ground displacement, outputting a real-time physical Factor of Safety (FS) for the terrain. Since a local hazard is only as dangerous as the asset's importance to the whole country, the engine computes the Exact Betweenness Centrality of the grid topology using NetworkX. This centrality acts as a threat multiplier, combining with local physical vulnerability to produce the final Systemic Risk Index (SRI). When a critical tower or substation exceeds the stability threshold, the system simulates its physical breakdown and instantly evaluates active energy pathways between regional power plants and substations, automatically compiling shortest-path re-routing protocols using NetworkX to bypass the bottleneck or triggering an emergency alert if the grid gets partitioned.Tech StackThe core engine runs on NumPy for matrix calculus and NetworkX for graph theory and centrality tracking. The interactive frontend is built with Streamlit to deliver a clean, high-fidelity engineering dashboard UI. Data visualization is handled via Plotly Graph Objects for a dynamic topological map with live network state rendering. The system also includes a native bilingual runtime engine via translations.py for Italian and English hot-swapping.