Power Quality Disturbance Detection and Classification using Deep Learning

Overview

This project presents a deep learning framework for automatic detection and classification of Power Quality Disturbances (PQD) using STFT spectrograms and Convolutional Neural Networks (CNNs).

Disturbance Classes

• Normal
• Voltage Sag
• Voltage Swell
• Harmonics
• Transient

Methodology

1. Synthetic Signal Generation
2. Noise Augmentation
3. STFT Spectrogram Generation
4. CNN-based Classification
5. Real-Time Inference

Technologies Used

• Python
• TensorFlow
• NumPy
• Matplotlib
• Signal Processing
• STFT
• Deep Learning
• CNN

Results

• Classification Accuracy: 95–98%
• Real-Time Inference Latency: <50 ms

Future Work

• Edge AI Deployment
• IoT Integration
• Wavelet Transform Based Analysis
• LSTM/GRU Models
• Real-World IEEE PQD Datasets

Project Status

Source code reconstruction is in progress. This repository currently contains the project documentation, methodology, architecture, and future implementation roadmap.