Anomaly Detection Technology for Gas-Insulated Switchgear

Condition monitoring technology has been established that digitizes analogue signals from gaspressure sensors, circuit-breaker stroke sensors, PD sensors and other sensors, and transmits them to higher-level servers, enabling sequential detection of failures. Furthermore, development is underway of pattern learning-based condition-monitoring techniques to detect precursors to abnormalities before they occur, enabling planned maintenance. The detection system divides waveforms into segments, computes a score for each segment representing the deviation between on-site waveforms and normal waveforms, and judges an abnormality when the score exceeds a threshold. For gas leaks and breaker operating characteristics, the system has been validated to avoid false positives under normal conditions and to detect simulated abnormal conditions.

For switching characteristics in particular, threshold settings are adjusted based on laboratory tests to account for effects that should not be attributed to equipment degradation, such as ambient temperature and control voltage, and for changes in switch behaviour after many normal operations.

Separately, although the adoption of dry air as an insulation medium is progressing as an SF6 alternative, the discharge characteristics of dry air are not yet well established. To address this,

PD characteristics were examined in high-pressure dry air and the PD sensor was developed, matched to those characteristics. To evaluate the performance of the sensor, effective height can be one indicator as the output voltage relative to the electric field strength at the sensor in addition to the detection sensitivity [1]. Test results showed the sensor’s lower frequency limit was extended from the conventional UHF band down to 100 MHz, achieving an effective height of 13 mm against a specified height of 6 mm and a detection sensitivity of 0.1 pC. For application to actual switchgear composed of multiple units, it is necessary to understand attenuation characteristics to determine the sensor location; investigations on this are ongoing.