Monitoring of Transformer Winding Looseness Based on Chaos and Clustering

In order to monitor the mechanical state of transformer winding more accurately and effectively, a new method based on phase space reconstruction and gray wolf optimization K-means (GWO-Kmeans) clustering algorithm is proposed. Firstly, according to the chaotic characteristics of transformer vibration signals, considering the visibility of reconstruction space, the embedding dimension is chosen to be 3. Secondly, the mutual information method is used to calculate the optimal delay time $\tau$ for phase space reconstruction of transformer vibration signal. Then, the gray wolf algorithm is used to optimize the K-means algorithm to select more reasonable initial cluster center, and then GWO-Kmeans algorithm is used to find the cluster center of reconstructed signal phase trajectory. Finally, according to the change of distance from the center displacement vector sum of cluster to the origin, the loose state of transformer winding is monitored. The results show that the GWO-Kmeans clustering algorithm effectively improves the accuracy of clustering results. The change of the center displacement vector and the distance between the coordinate and the origin of the phase trace cluster of the transformer vibration signal can reflect the loose state of the winding, thus providing a theoretical basis for the maintenance of the loose state of the transformer winding.