GIS internal fault diagnostics using artificial neural networks

This panel describes the application of ANN to internal fault detection of GIS (gas insulated switchgear). The goal of this application is the predictive maintenance of the system. The GIS is monitored on-line through attached-sensors to detect small symptoms of abnormalities before a fatal malfunction. A new method of ANN architecture called ICLNN (incremental cluster learning neural network) is employed to perform recognition of patterns to the averaged spectrum of sensor signals. The working of the prototype system is demonstrated with some experimental results to illustrate the advantages of the ANN