A Complete EEMD with Adaptive Noise Analysis and Improved LSSVM for Transformer Winding Looseness Fault Diagnosis

To address the difficulty of identifying transformer winding looseness faults, a transformer winding looseness fault diagnosis approach based on complete ensemble empirical mode decomposition with adaptive noise analysis (CEEMDAN) and improved whale optimization algorithm optimized least squares support vector machine (IWOA-LSSVM) is come up. Firstly, CEEEMDAN is employed to process the collected transformer vibration signal to achieve modal components IMFs. Secondly, this study creams off the IMF with the optimal correlation and count its multi-scale arrangement entropy to form a feature dataset. Then, the IWOA is employed to make the most of LSSVM to construct the IWOA-LSSVM blueprint. The experimental results confirm that the suggested approach can triumphantly diagnose transformer winding looseness faults, and could provide assistance for live maintenance of transformers.