Switchgear Insulation Condition Assessment with Multi-Dimensional Feature Database Based on an Optimal Grade Clustering Method

In order to evaluate a switchgear insulation condition assessment with live detection data more objectively and effectively, an optimal grade clustering method with a multi-dimensional feature database is proposed. This paper first analyzes the mechanism of partial discharge (PD), and volatility indicators and a life factor are introduced to fully exploit the characteristics of PD, and are used to establish the multi-dimensional feature database. Using the established database, the proposed K-means clustering algorithm, which includes data cleaning and optimal grade, is presented to classify the switchgear insulation condition. The cluster-outlier algorithm with relative distance (RD) is used to realize data cleaning. The optimal grade is designed to solve the problem of picking the number of clusters subjectively by means of the sum of the squared errors (SSE). The feasibility of the algorithm is verified via live detection data, which provides a certain theoretical basis for the switchgear insulation assessment.