Effect of noise on the location accuracy of partial discharges using radiated RF detection techniques

Partial discharges (PD) are electric discharges that occur when the insulation system of high voltage plant is in the process of degradation. Prior to complete breakdown, PDs provide an early warning of imminent equipment failure, while also accelerating the breakdown process. Radiated radio frequency (RF) measurement methods are one of numerous other detection methods which may be employed to locate PD sources. This method uses at least 4 receiving antennas to detect the radiated PD RF signals from which the three dimensional position of the PD source can be calculated using iterative or non-iterative location algorithms. This paper evaluates the effect of Gaussian noise on the location accuracy of two different PD sources which have different frequency spectra and signal strength characteristics using a locally customised "Y shaped" arrangement of four antenna positions. Based on measurements of hundreds of experimental RF signals, the time-differences-of-arrival (TDOA) between the different antennas is calculated using cumulative energy and then the location of the PD source determined by applying the iterative hyperbolic least squares (HLS) algorithm. The results demonstrate that the accuracy and the ability to determine the position of the two different PD sources depends on the signal-to-noise ratio (SNR) quality as well as the strength of the signals. Also, results show that averaging the TDOAs may not provide a converged solution, while averaging the locations of the individual TDOAs provide a better solution.