Spatio-Temporal Fault Contagion Tracking and Predictive Asset Management for Offshore Wind Farms Based on Multi-Terminal Harmonic Fingerprints

Project Identity: Shinar of Clark
Author: Yi Zeng
Framework: Spatio-Temporal Fault Contagion Tracking and Predictive Asset Management
📄 Manuscript PDF: Read the Full Paper Here
DOI: https://doi.org/10.5281/zenodo.20306854
github:https://github.com/Shinar-of-Clark/Clark-Paradigm-Initiative
Email:Clark@ShinarOfClark.com

📖 Introduction

This repository implements Phase 3 (“Pathology & Prognosis”) of the sub-health diagnostic framework for offshore wind assets under the “Clark Paradigm”.

Building upon the “Electromagnetic Ledger” (Phase 1) and the “Proactive Protection Shield” (Phase 2), this phase focuses on cross-terminal fault contagion tracking and proactive asset management. By tracking the spatio-temporal propagation of sub-health anomalies across a “5-Level 7-Node” distributed topology, this phase achieves a critical transition from localized anomaly detection to network-wide cascading propagation analysis and Remaining Useful Life (RUL) estimation.

Core Philosophy: Fault Contagion Tracking = Spatio-Temporal Anomaly Envelopes + Cascade Propagation Time-Delay Prediction (Δt).

📁 Repository Structure

To ensure the transparency and reproducibility of our research, the supporting materials for this paper are organized as follows:

  • figures/: Contains all high-resolution figures, attention heatmaps, and fault contagion mechanism flowcharts presented in the manuscript.
  • scripts/: Contains Python implementation scripts, including the CNN feature trigger, the Transformer spatio-temporal sequence analyzer, and figure generation routines.
  • data/: Contains the generated simulation datasets representing multi-terminal harmonic signals under different fault propagation scenarios.

🛠️ Technical Architecture

This phase implements a hybrid deep learning framework to track how sub-health states propagate across the multi-terminal electrical layout:

  • 5-Level 7-Node Spatio-Temporal Topology: Maps offshore wind assets into WTG (Level A/Node A), Converter (Level B/Node B), Collector Cables (Level C/Nodes C1, C2), Transmission Cables (Level D/Nodes D1, D2), and Point of Common Coupling (Level E/Node E).
  • Two-Stage Collaborative Model:
    • Stage 1 (CNN Anomaly Envelope Trigger): Translates high-dimensional frequency-domain harmonic signals (Xt∈R7×50) into real-time anomaly probabilities (pt).
    • Stage 2 (Transformer Spatio-Temporal Analyzer): Uses multi-head self-attention with spatio-temporal positional encoding to track the propagation path and estimate the contagion time-delay (Δt).

🚀 Key Features

  • Spatio-Temporal Contagion Tracking: Traces how transient anomalies and sub-health degradation propagate downstream (from turbine to grid) or upstream (from grid disturbances back to the turbine).
  • CNN Anomaly Trigger: Captures early fault envelopes with high robustness against measurement noise and fluctuating environmental conditions.
  • Transformer-based Delay Estimation (Δt): Quantifies the contagion propagation delays, allowing O&M systems to identify the root cause and forecast warning lead times.
  • Multi-Terminal Harmonic Fingerprinting: Relies on low- and mid-frequency harmonic envelopes (up to the 50th order) to identify mechanical and electrical issues (e.g., blade icing, cable degradation) without expensive high-frequency sensors.

📊 Performance

By utilizing the collaborative CNN-Transformer model, the platform achieves high accuracy in predicting cross-terminal fault contagion delays:

  • Cascade Tracking: Successfully forecasts downstream propagation paths and provides early warning windows of 10 to 30 seconds prior to catastrophic grid shock or terminal failures.
  • Interpretability: The self-attention weight heatmaps clearly highlight the causal dependencies and bottleneck nodes in the 5-Level 7-Node architecture.

📚 Citation

If you utilize the concepts or content of this project in your research, please cite our manuscript:

APA Format:

Zeng, Y. (2026). Spatio-Temporal Fault Contagion Tracking and Predictive Asset Management for Offshore Wind Farms Based on Multi-Terminal Harmonic Fingerprints (v1.0.0). Zenodo. https://doi.org/10.5281/zenodo.20306854

BibTeX:

@misc{zeng2026clark_contagion,
  title={Spatio-Temporal Fault Contagion Tracking and Predictive Asset Management for Offshore Wind Farms Based on Multi-Terminal Harmonic Fingerprints},
  author={Yi Zeng},
  year={2026},
  publisher={Zenodo},
  version={v1.0.0},
  doi={10.5281/zenodo.20306854},
  url={https://doi.org/10.5281/zenodo.20306854}
}

🛡️ License

This project is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) License.

License: CC BY 4.0

Rights Statement: You are free to share and adapt this work, provided that you give appropriate credit to the author Yi Zeng (Project Shinar of Clark) and indicate if changes were made.

滚动至顶部