Detection of Partial Discharge Anomaly in Switchgear Based on Principal Component Clustering Outlier Algorithm

Based on the on-line detection data of the switchgear, a principal component analysis clustering outlier algorithm based on multi-dimensional feature quantity is proposed to identify insulation state of the switchgear. Firstly, we use the life cycle coefficient of the switchgear and the dispersion of the partial discharge detection data, the average distance percentage and the maximum volatility to construct the principal component of the multi-component sample database, which is comprehensively quantified the deterioration degree of the insulation state. Then choosing the optimal cluster parameters of outlier algorithm by Silhouette Coefficient; Considering the density difference between the clusters, the relative distance is proposed to quantify the degree of deterioration of the insulation state to realize the abnormal detection of insulation state of the switchgear. The case analysis of the living detection data is carried out to verify the feasibility of the method, which provides a theoretical basis for the abnormal identification of the insulation state to the switchgear.