TyMill/FedGridSim: v1.2.0

Abstract

FedGridSim v1.2.0 — Reviewer Revision Release Overview FedGridSim v1.2.0 is the reproducibility release accompanying the revised manuscript: "Federated Learning for Resilient Smart-Grid Monitoring: A Physics-Based Simulation Framework for Distributed Energy Networks." The release consolidates the simulation, federated-learning, robustness, contingency-screening, and evaluation components used in the study and aligns the public software version with the methodology reported in the revised manuscript. What is new in v1.2.0 This release focuses on methodological clarification, reproducibility, and additional reviewer-requested validation. Key updates include: clarified positioning of FedGridSim as a physics-based AC power-flow simulation framework for distributed smart-grid monitoring; explicit and reproducible N-1 and N-2 generator contingency screening based on a fixed stress-ranking procedure; improved documentation of the simulated closed-loop false-data-injection (FDI) scenario and its effect on both observed telemetry and the underlying physical operating state; clearer separation between: observed client telemetry, physics-derived grid quantities, prediction targets, hidden audit variables used only for validation; clarified interpretation of the hourly temporal resolution as a supervisory / operational energy-management timescale rather than protection- or PMU-level dynamics; refined description of the study-specific validation-blended personalized FedAvg baseline; explicit accounting of federated model-update traffic and centralized data-transfer traffic; added reviewer-driven operational-threshold sensitivity analysis for congestion, overload, and voltage-instability definitions; improved reproducibility metadata and alignment between the manuscript, repository, experiment manifests, and archived release. Threshold-Sensitivity Validation The additional sensitivity campaign evaluates reasonable perturbations of the operational thresholds used to define grid-risk labels. The analysis covers: congestion threshold: 85%, 90%, 95%; overload threshold: 95%, 100%, 105%; voltage-violation tolerance: 0.0025, 0.0050, 0.0075 p.u.; severe voltage violation: 0.015, 0.020, 0.025 p.u.; voltage persistence: 1, 2, 3 consecutive steps. The targeted IEEE 118-bus N-1 experiment uses: 20 independent physical seeds; 5 primary comparison methods; 15 threshold configurations; 1500 model fits in total. The qualitative separation between federated and non-federated approaches remains stable across the tested threshold range, although the exact ordering among federated variants remains metric- and threshold-dependent. Reproducibility Scope The release supports experiments based on IEEE benchmark systems and AC power-flow simulation using pandapower. The study evaluates: IEEE 14-bus, IEEE 39-bus, IEEE 118-bus, under normal operation and multiple physical/cyber-physical stress scenarios, including generator outages, N-2 generator outages, demand and renewable stress, telemetry corruption, and closed-loop FDI. The implementation is intended for simulation-based research and reproducible benchmarking. It does not represent a hardware-in-the-loop, SCADA, PMU, or field-deployed smart-grid testbed. Citation If you use FedGridSim, please cite the archived Zenodo record: Miller, T. I. FedGridSim v1.2.0. Zenodo. https://doi.org/10.5281/zenodo.21889600 Zenodo record: https://zenodo.org/records/21889600 Repository GitHub repository: https://github.com/TyMill/FedGridSim Version FedGridSim v1.2.0