A Fault Identification Method for Ferroresonance Based on a Gramian Angular Summation Field and an Improved Cloud Model

Due to the broad frequency domain and nonlinear characteristics of ferroresonance signals, traditional time–frequency analysis methods often face challenges such as misjudgment, difficulty in threshold setting, and noise interference when extracting features from ferroresonance overvoltage signals. A fault identification method for ferroresonance based on the Gramian Angular Summation Field (GASF) and an improved cloud model is proposed to address the identified problems. Firstly, this paper employs Symplectic Geometric Mode Decomposition (SGMD) to denoise the ferroresonance overvoltage signal, extract its characteristic modal components, and reconstruct the signal. Secondly, the reconstructed one-dimensional signal is transformed into a two-dimensional image using GASF. Subsequently, we extract texture features of GASF images with different resonance types by grey-level co-occurrence matrix (GLCM) and establish the corresponding cloud distribution model to characterize these textures. Finally, we calculate the membership degree between the standard cloud for the signal to be identified and the index cloud in the cloud distribution model, enabling accurate identification of the type of ferroresonance based on this membership degree. Simulation and actual measurement data analyses validate the feasibility and effectiveness of the proposed method.