Machine Learning-Based Real-Time Power Quality Disturbance Classification using Kalman Filter Segmentation and Deep Learning

IEEE PES Innovative Smart Grid Technologies Conference Europe (ISGT Europe), pp. 1–5

Abstract

The growing complexity of electrical grids necessitates robust methods for monitoring power quality (PQ) disturbances in real-time. This work presents a complete framework for PQ classification that integrates signal processing, intelligent segmentation, and machine learning models. Central to this framework is the application of a Kalman Filter (KF), which serves as an adaptive event detection and signal segmentation tool, enabling the isolation of transient and non-stationary events from continuous voltage waveforms. Four feature extraction techniques—FFT, STFT, WT, and ST—are combined with Support Vector Machine (SVM) classifiers and benchmarked against a Convolutional Neural Network (CNN) operating on Hilbert-transformed signal envelopes. Results demonstrate that the CNN, combined with Kalman-filter-based segmentation, outperforms traditional methods both in accuracy and in computational efficiency for real-time deployment. This highlights the practical viability of integrating adaptive filtering with AI for future smart grid monitoring systems.

Authors 4

1. Markus Stroot Aachen

   Fraunhofer Institute for Applied Information Technology · RWTH Aachen University

   Affiliation as printed

   IAEW at RWTH Aachen and Fraunhofer FIT,Aachen,Germany

2. Rui Wang Aachen

   RWTH Aachen University

   Affiliation as printed

   IAEW at RWTH Aachen University,Aachen,Germany

3. Amirali Mahjoob Aachen

   RWTH Aachen University

   Affiliation as printed

   IAEW at RWTH Aachen University,Aachen,Germany

4. Andreas Ulbig Aachen

   Fraunhofer Institute for Applied Information Technology · RWTH Aachen University

   Affiliation as printed

   IAEW at RWTH Aachen and Fraunhofer FIT,Aachen,Germany