Detecting partial discharge by AI approach

Abstract

This project investigates the efficacy of Artificial Intelligence (AI) models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), and Support Vector Machines (SVMs), for detecting Partial Discharge (PD) in electrical systems using waveform data. Key to our approach was cleaning the dataset through denoising, standardization, and advanced feature extraction techniques like Fast Fourier Transform (FFT) and Continuous Wavelet Transform (CWT). Our results highlight the trade-offs between the models in terms of processing speed and temporal analysis capabilities. We also explored model deployment on portable devices, identifying significant challenges related to computational resource constraints. Future work will focus on data augmentation to simulate real-world signal characteristics and algorithmic improvements to optimize model performance for practical PD detection applications.