Research on fine-grained assessment of low-voltage line voltage quality based on GAN and parameter identification

The deep application of intelligent algorithms in voltage quality analysis has advanced the refined management of low-voltage distribution systems. Addressing the challenge of traditional methods in characterizing non-stationary signal features, this study constructs a voltage quality assessment model integrating Generative Adversarial Networks (GANs) with parameter identification. High-dimensional input optimization is achieved through Energy Dispersion Modulation (EMD) decomposition and energy entropy feature extraction. The stability of generated features is enhanced by combining the Wasserstein loss function with a gradient penalty mechanism. Experimental results demonstrate that this model achieves approximately 14% higher evaluation accuracy than traditional algorithms across multiple scenarios, exhibiting superior convergence performance and real-time capability.