Application of a Partial Discharge Diagnosis Method Based on the Novel Multispectral Array Sensor and GMM in Different Insulating Gases

Optical detection of partial discharge (PD) is an important means to diagnosis the insulation status of equipment. However, the current optical detection is either unable to perform simultaneous detection of multispectral features or it can only be applied under laboratory conditions. Moreover, the PD of different gases will correspond to different multispectral characteristic distributions, but the current research is only aimed at the study of SF
₆
. Therefore, this article proposes a gas-insulated equipment PD detection method based on the multispectral array sensor and Gaussian mixture model (GMM) clustering algorithm in C
₄
F
₇
N/CO
₂
gas mixture and SF
₆
. This method adopts the silicon photomultiplier (SiPM) array and multispectral grid to realize the collection of PD optical signals of different wavelength bands, which has a high degree of integration and can be applied in actual equipment. Then through the GMM model, cluster analysis of different multispectral features can effectively diagnose different types of PD defects. In the experiment, four kinds of PD defects under the conditions of five various ratios C
₄
F
₇
N/CO
₂
gas mixture and pure SF
₆
gas are used for test. The similarities and differences of the PD multispectral characteristics in phase distribution, energy distribution, and radar chart distribution are studied, which compares the PD multispectral characteristics in C
₄
F
₇
N/CO
₂
gas mixture and SF
₆
. According to the experiment result, the diagnosis accuracy in different gases is generally higher than 84%, and the highest can reach 95% in SF
₆
. This result shows that the method can well diagnose the PD in both SF
₆
and C
₄
F
₇
N/CO
₂
power equipment, which has a wide range of application prospects.