Power Transformer Fault Diagnosis using Dynamic Multiscale Graph Modeling and M2SGCN Network Based on Statistical Fusion

Abstract—Power equipment fault diagnostics hold significant importance for the stability of power grid systems. In pursuit of this objective, this paper proposes a fault diagnosis method that utilizes dynamic multiscale graph modeling and the M2SGCN network, incorporating statistical fusion. Specifically, we propose a novel dynamic multiscale graph (DMG) modeling method to derive visibility graph (VG) data and horizontal visibility graph (HVG) data from vibration signals across multiple scales. Next, we establish M2SGCN, a comprehensive neural network architecture that learns global and local features simultaneously, providing a more precise representation. Finally, we utilize a Dempster Shafer evidence theory statistical fusion technique combined with an adaptive threshold model (DSTFusion) to integrate primary decision results for enhanced fault diagnosis accuracy. Furthermore, we analyze two datasets obtained from single-phase and three-phase power transformers to demonstrate the evolution process. When compared to state-of-the-art indicators such as accuracy, precision, recall, and F1-scores, the proposed method excels in multiple aspects, successfully detecting fault states prior to their occurrence and achieving outstanding performance.