⚡ HE-PDA (High-Frequency Edge-Side Partial Discharge Analyzer)

Edge-Based High-Frequency Partial Discharge Analysis

HE-PDA is a professional diagnostic platform for high-frequency transient signal processing and Partial Discharge (PD) analysis. By moving complex feature extraction directly to the edge, it converts high-frequency physical phenomena into automated, actionable insights. It provides a robust, efficient solution for assessing the insulation health of high-voltage cables and power equipment.

▲ Phase 0 View: Hardware-accelerated PRPD spectrum and expert diagnostics.


▲ Phase 1 View: Tracing micro-features across different insulation states.


▲ Phase 2 View: Background noise stripping and interference analysis.

✨ Key Features

• 🚀 Efficient Edge-Side Feature Extraction (OOM-Resistant): Proven at sampling rates from 500KHz to 80MHz (HF/VHF). The "dimensionless filtering" architecture is built to scale up to UHF (300MHz+). By using SOS-cascaded high-pass filters and adaptive max-pooling, the system converts massive raw datasets into sparse PD events while keeping memory usage extremely low—making it ideal for constrained hardware like the STM32.
• 📊 Hardware-Accelerated Visualization: Powered by Dash and Plotly WebGL, the platform provides lag-free rendering for 100,000+ data points (50-cycle accumulation). Features include multi-cycle dynamic drill-down and microscopic evolution tracing for PRPD patterns.
• 🧠 Explainable Expert Diagnostics: Built on IEEE 1434 and CIGRE standards, our system moves beyond "black box" models. It uses 6-sigma dynamic noise stripping and decision trees to reliably identify internal voids, surface tracking, floating potentials, and inverter-switching noise.
• 📍 Intelligent TDR Echo Location: Automatically detects dual-end reflected waves to calculate defect locations and provides a reliability score based on phase-locking status.
• 🐳 One-Click Reporting & Deployment: Export full diagnostic reports (Diagnostic_Report.html) instantly. Includes a native Dockerfile for containerized deployment on Linux/Windows servers or edge gateways.

📦 Ready-to-Use Datasets

The platform includes built-in high-frequency samples for immediate testing. For deep validation or secondary development, we provide access to 8,000+ sets of high-quality PD waveforms (80MHz, single-cycle):

• 🔗 Kaggle Official Dataset (VSB Power Line Fault Detection)
• 🔗 Author's Google Drive Backup

Update History

v1.0.0

• 🚀 PDA Diagnosis Engine - Version 1.0.0: First stable release of the Partial Discharge Diagnosis Engine (PDA).
• Standalone Binary: Fully self-contained execution; no Python environment or library installation is required on the host system. Built using Nuitka.
• Advanced Signal Processing: Integrated with optimized scipy and numpy stacks for high-frequency partial discharge (PD) signal analysis and filtering.
• Interactive Visualization: Built with Dash and Plotly to provide a real-time, responsive diagnostic dashboard.
• Production Ready: Specifically tuned for 1GB RAM Ubuntu 22.04 VPS environments with a focus on memory stability.
• Documentation Update: Updated the Quick Start guide to reflect the new standalone binary release deployment process.

🚀 Quick Start

🛠 Deployment Instructions (Ubuntu 22.04+)

1. Download the he_pda_engine from the Release Assets section.
2. Grant execution permissions:

   chmod +x he_pda_engine

3. Run the engine:
   • For direct testing:

     ./he_pda_engine

   • For background production (recommended):

     nohup ./he_pda_engine > app.log 2>&1 &

📝 Notes:

• The default service port is 8052. Please ensure this port is open in your VPS firewall/security group settings.
• If you encounter a libglib error on a fresh Ubuntu install, run:

  sudo apt update && sudo apt install -y libglib2.0-0