A multi-modal deep learning framework for power quality disturbance classification: An integration of 1D time-series signals and 2D scalograms

Abstract

Power quality (PQ) is crucial for the dependable functioning of electrical systems, requiring stable voltage, frequency, and waveform integrity. However, power quality disturbances (PQDs), resulting from faults, nonlinear loads, and switching events, can degrade system performance, damage equipment, and reduce operational efficiency. Accurate identification and classification of PQDs are therefore critical for effective mitigation. Traditional methods that rely solely on either one-dimensional (1D) time-series signals or two-dimensional (2D) waveform images often fail to capture the full characteristics of disturbances, leading to reduced accuracy. To address this limitation, a multi-modal deep learning framework is proposed that integrates 1D time-series data with corresponding 2D scalogram images. The proposed model employs parallel 1D and 2D convolutional neural networks (CNNs), each enhanced with attention mechanisms to enhance feature extraction by focusing on modality-specific salient information. The proposed model is evaluated on a comprehensive synthetic dataset of sixteen PQD types. Experimental results demonstrate that the proposed approach achieves an average classification accuracy of 99.99%, a sensitivity of 99.98%, and a specificity of 99.99%, outperforming existing methods. These results demonstrate the framework’s robustness and its potential as an effective solution for PQD monitoring and classification in smart grid environments.