A New Precise Power Quality Disturbance Identification Framework Based on Two-Dimensional Characterization Feature Enhancement and Deep Learning

The increasing integration of renewable energy sources into electrical grids has exacerbated power quality issues, necessitating advanced methods for the rapid detection and precise classification of power quality disturbances (PQDs). This study presents a novel PQD identification approach that integrates two-dimensional feature enhancement with a deep learning framework to address these challenges. The proposed method employs the relative position matrix (RPM) technique to transform PQD signals into visual representations, enhancing 2D feature extraction by capturing temporal dependencies and inter-point relationships through spatial arrangement. Building on this, Spatial Group-wise Enhance (SGE)-MobileViT, an advanced identification and classification technique that autonomously extracts image features, was introduced for accurate PQD detection. The SGE-MobileViT model incorporates an attention mechanism that adaptively adjusts the feature map significance, optimizing feature space scalability and enabling the effective capture of both local features and global contextual relationships. Experimental results demonstrated the model’s superior performance, achieving 99.17% classification accuracy in noiseless environments and maintaining high accuracy (95.13%, 97.00%, and 97.50%) at signal-to-noise ratios of 20 dB, 30 dB, and 50 dB, respectively. The robustness and practical applicability of SGE-MobileViT were further validated through comprehensive simulations and hardware platform implementations including an embedded system demonstration. This study offers a significant advancement in PQD identification, providing a reliable solution for power quality management in modern electrical grids with high renewable energy penetration.