Multi-Parameter Fusion-Based Operational State Assessment for HV GIS Disconnector Mechanisms

To address the Operational State Assessment of high-voltage gas-insulated switchgear (GIS) disconnectors, this paper proposes a novel assessment method integrating mechanical thermal electrical multi-parameter fusion with a Temporal Convolutional Network combined with a Deep Ordinal Constraint (TCN-DOC) model. Based on a comprehensive 220 kV GIS experimental platform, the proposed method synchronously extracts and fuses features from vibration, temperature, and partial discharge signals under typical fault conditions. By introducing deep ordinal constraints into the loss function, the TCN-DOC model optimizes the feature space to be “compact intra-class, orderly inter-class,” achieving a high test-set accuracy of 96.24% and significantly outperforming traditional deep learning models. Furthermore, an intelligent operation and maintenance decision-making process based on the Health Index (HI) was designed. This study provides a complete technical framework and theoretical reference for the precise condition-based maintenance of GIS equipment and the safe operation of power grids.