Research Description
Early identification of partial discharges (PD), which represent key indicators of insulation degradation, is crucial to averting equipment failures and enhancing overall system dependability. This paper offers a thorough examination of the use of Self-Organizing Maps (SOMs) for PD-based condition monitoring in high voltage (HV) substations, with particular emphasis on Air Insulated Switchgear (AIS) and Gas Insulated Switchgear (GIS). Using simulated PD waveforms generated in MATLAB, an 8×6 SOM grid effectively clustered three discharge types: corona, surface, and internal. Performance metrics demonstrate high classification accuracy, with F1-scores of 0.8586, 0.9249, and 0.9899 respectively. The low AQE (0.0535) and clear 3D visualizations confirm SOM's capability to discriminate PD patterns and to clearly separate AIS and GIS characteristics. The proposed SOM framework highlights strong adaptability, interpretability, and scalability for real-time PD analysis. Results demonstrate the great potential of SOM-based monitoring to support predictive maintenance and enhance the reliability of modern power substations.
