kalamangna/1d-cnn-fault-detection

1D-CNN Power System Fault Detector

An intelligent, end-to-end system for real-time detection, classification, and localization of faults in electrical transmission lines using Deep Learning.

🚀 Key Features

• —Multi-Task Learning: Simultaneous Prediction of Status (Detection), Type (Classification), and Jarak (Localization mapped to 5 discrete points: 0.25, 1.25, 2.5, 3.75, 4.9 km).
• —High-Resolution Signal Analysis: Processes 3-phase Voltage and Current signals at 10kHz.
• —Modern Dashboard: Built with Nuxt 4, featuring real-time signal visualization in both per-unit and physical units.
• —Automated Pipeline: Full lifecycle support from raw Excel data processing to model deployment.

📂 Architecture

• —Backend (Flask): Serving a 1D-Convolutional Neural Network (1D-CNN) trained on multi-task objectives.
• —Frontend (Nuxt 4): Industrial-grade UI for signal ingestion, analysis, and metric reporting.
• —AI Model: Multi-output architecture optimized for transmission lines up to 5.0 km.

📂 Project Structure

• —`/backend`: Python Flask API, Signal processing scripts, and Docker configuration.
• —`/frontend`: Nuxt 4 application, Vercel config, and Signal visualization hub.
• —`/backend/artifacts`: Model weights (.keras), Scalers, and detailed Evaluation plots (Confusion Matrices, Training Curves).

🐳 Deployment (Production)

Docker Compose (Recommended)

You can deploy the entire stack using Docker:

docker-compose up -d --build

This will launch the backend on port 5001 and the frontend on port 3000.

Vercel (Frontend only)

The frontend is optimized for Vercel deployment. Ensure you set the NUXT_PUBLIC_API_BASE_URL environment variable to point to your deployed backend.

📊 Model Performance & Testing

• —Verified Metrics: Achieved 100% Detection Accuracy, 100% Classification Accuracy, and a 0.23 km Location RMSE on a 5.0 km transmission line.
• —Metrics Dashboard: View detailed Accuracy, Precision, Recall, and F1-Score along with Training Curves and Confusion Matrices.
• —Lab (Analysis): Test the model using explicit "Normal" or "Fault" data generation, or by uploading custom CSV data.
• —Inception Detection: Real-time calculation of fault inception time (MS) based on voltage sag (V < 0.85) and current spike (I > 0.3) analysis.
• —Automated Tests: Run backend logic verification using unittest:

  cd backend && venv/bin/python3 -m unittest tests/test_logic.py

🛠 Tech Stack

• —AI/ML: TensorFlow 2.16+, Scikit-learn, Pandas, NumPy.
• —Backend: Flask, Flask-CORS, Joblib.
• —Frontend: Nuxt 4, Vue 3, Tailwind CSS, Chart.js.

⚡ Quick Start

1. Backend Setup

cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python app.py

2. Frontend Setup

cd frontend
npm install
npm run dev

📖 Documentation

Detailed technical documentation and user guides are available in:

• —frontend/public/docs/pipeline.pdf
• —frontend/public/docs/documentation.pdf