Abstract: Sensitive signal perception and precise fault diagnosis are the key to the application of optical detection for partial discharge. In this paper, a micro-sensing device with multiple independent spectral channels was developed based on the single-photon solid state optoelectronic devices. Then, a multi-spectral synchronous measurement of typical partial discharge in SF₆ was carried out based on this device, and the optical pulse waveform, pulse train, phase-based statistical map and other aspects were analyzed. Furthermore, a multispectral data analysis method based on the ternary graph was proposed, demonstrating the advantages of multispectral partial discharge diagnosis technology. Finally, the possible fusion method of GIS was proposed. The results show that the micro multispectral sensor can effectively detect partial discharge light pulses in different wavebands; the intensity and proportion of the multispectral pulses have a strong correlation with partial discharges' types. It shows obvious statistical rules on the statistical and ternary map. Compared with traditional partial discharge diagnosis methods, multispectral detection can provide higher-dimensional discharge information, and a more abundant and accurate basis for fault recognition and risk assessment, which is not affected by the location of discharge.