Partial Discharge Positioning Based on Statistical Analysis of UHF RSSI Values

Partial Discharge (PD) detection and localization is an important indicator for the safe and stable operation of power equipment. Unlike the commonly used time-difference of UHF signal method for PD localization, the method based on the Received Signal Strength Indicator (RSSI) was proposed due to its low hardware requirement. However, exist RSSI methods are mainly focus on the fingerprint-based technology, which is hard to performed because of the prebuilt of fingerprint map is a heavy burden in site environment. Therefore, this paper proposes a RSSI data machine learning scheme for PD localization. The Gaussian Process Classification (GPC) algorithm is used to mitigate the multipathing and shadowing effects. Besides, a Maximum Likelihood Estimation (MLE) analysis method of RSSI data is also applied for PD localization. The experimental test is performed and the results indicate that the mean error of PD localization is about 0.378 meters. From practical application, it satisfies the requirement to distinguish the power equipment that having insulation deterioration.