Transformer Internal Mechanical Fault Diagnosis Method Based on Improved CEEMDAN and IMVO-HKELM

Aiming at the problem that it is difficult to identify the internal mechanical fault of transformer accurately and effectively, a method of transformer internal mechanical fault diagnosis based on improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and improved multi-verse optimization algorithm optimized hybrid kernel based extreme learning machine (IMVO-HKELM) is proposed. Firstly, to avoid the generation of false modal components, an improved ICEEMDAN is used to decompose the original vibration signal of the transformer. Secondly, the Pearson correlation coefficient method is used to select the modal component with the highest correlation and calculate its MFE value. Then, the MFE value is used as the feature quantity to construct the feature dataset, and the core parameters and regularization coefficients of HKELM are optimized using the spiral update and adaptive compressibility factor improved MVO. Then, the MFE value is used as the feature quantity to construct the feature dataset, and the HKELM kernel parameters and regularization coefficients are optimized by using the spiral update and adaptive compressibility factor improved MVO. Finally, the feature dataset is input into the established IMVO-HKELM model for classification and recognition. The experimental results indicate that the proposed method can accurately diagnose mechanical faults of transformer winding and iron core looseness, with a diagnostic accuracy of over 99%, and can provide certain guidance for the formulation of on-site maintenance strategies for transformers.