Partial discharge (PD) detection and classification is important to avoid early failure of gas insulated system (GIS). Ultra high frequency (UHF) sensing technique is popular for PDs detection in GIS due to its ability to work in all environmental conditions, and its detectability to all types of PDs. Lot of UHF sensors were employed for PDs detection for many years but, their performances were not sufficient. The aperture sizes of these antennas are large, and they work in narrow band or multiple bands in UHF spectrum. They also suffer from low gains, and are not covered with protective cover which is important to operate in high voltage (HV) environment. Hence, it is essential to develop compact, reliable UHF sensors for detection of partial discharges inside GIS. Besides, the propagation characteristics of PD signals inside GIS were affected due to internal components, electrical discontinuities and branching. Many studies on the UHF signal propagation in GIS are limited to numerical investigations with varying GIS geometry, defect location and UHF sensor orientation with the use of numerical probes. But the effect of defect size and rise time of PD pulse are not studied. So, the complete study of the PD induced UHF signal characteristics inside GIS with designed UHF sensors is important to determine the PD detection capability and optimal placement of the UHF sensors inside the GIS, prior to field testing. Further, the sensitivity of the detected PD signals inside GIS depends on the sensor’s polarization, which needs to be studied in detail.
In this work, three UHF sensors were developed with varying sizes and characteristics. Sensor1 is a conical monopole antenna (CMA), sensor2 is a tapered planar spiral antenna (TPS) and sensor3 is cosine slot Archimedean spiral antenna (CSASA). All the sensors were designed with finite element method (FEM) based electromagnetic simulation software, Ansys HFSS®. The sensors were fabricated; and characterized using a vector network analyser (VNA). CMA sensor operates from 0.5 to 3 GHz with peak gain of 1 dB, and has aperture diameter of 150 mm. While, the TPS sensor operates from 0.5 to 3.5 GHz with peak gain of 1.68 dB and its aperture diameter is 159 mm. The CSASA sensor operates from 0.5 to 5 GHz with peak gain of 4.95 dB and consists of 158 mm aperture diameter. Besides, CMA sensor is linearly polarized (LP), while the TPS and CSASA sensors are circularly polarized (CP).
A detailed study to understand the influence of PD defect parameters such as location, orientation, size and rise time on the propagated signals in GIS were studied inside a 3 meter long L shaped GIS (L-GIS) section using LP CMA sensors. The analysis was performed in both transient and steady state simulations, experimentally verified in a laboratory L-GIS setup. Profound influence of these defect parameters on the propagated PD induced UHF signals inside GIS was observed. The effect of sensor’s polarization on the detected PD induced UHF signals was studied experimentally using LP CMA and CP TPS sensor inside L-GIS section. It was found that the circular polarization of the UHF sensor was advantageous in detecting the PD signals induced due to defects that were located in any orientation w.r.t UHF sensor.
Finally, all the sensors performance for detecting various types of PD defects were tested in laboratory L-GIS setup. The sensors were able to detect PD signals due to several types of defects inside L-GIS with signal strength higher than commercial Disc sensor. The circularly polarized high gain CSASA sensor demonstrated the higher PD signal strength when compared to other sensor’s signals. The PD signals that were detected before L bend and after L bend of L-GIS due to 7 types of PD defects were classified using convolutional neural network. The defects were effectively classified effectively with better than 98% accuracy.
Keywords: Classification, gas insulated system, high voltage, partial discharges, UHF sensors