UHF Partial Discharge Localization Methodology Based on Generalized Regression Neural Network

Abstract

Partial discharge (PD) detection and localization is an important issue in condition monitoring and diagnosis of power equipment. The existing Ultra-High Frequency (UHF) PD localization methods are mainly based on time difference technique which is bearing a high cost. Therefore, a low-cost PD localization method based on generalized regression neural network and received signal strength indicator (RSSI) fingerprint is proposed in this paper. The proposed method consists of two stages. First, the raw UHF RSSI data is collected by a wireless UHF sensor array to build the RSSI fingerprint map. Second, the online RSSI data of PD are measured and the position of PD source is calculated by a generalized regression neural network. The field test shows that the average localization error of our proposed method is O.51m, and 81.6% of localization errors are less than 1m. For RSSI based positioning systems, a lower bound on the mean square error (MSE) of a position estimate can be calculated by the Cramer-Ran Lower Bound (CRLB). The Log Normal Shadowing Model is used to calculate the CRLB in this paper, and 66.7% of PD localization results obtained by the proposed method are better than the results of Log Normal Shadowing Model.