Locating Partial Discharge Using Particle Swarm Optimisation

摘要

The use of radio frequency measurement (particularly at UHF) as a means of detecting, diagnosing and locating Partial Discharges in high voltage equipment has advanced considerably in recent years. Partial Discharge location based on UHF signals uses the time-difference-of-arrival of the received pulses at a number of sensors positioned around the equipment exhibiting internal Partial Discharge or arcing in order to establish the location of its source. This method of locating the source is complicated when the line-of-sight path from the Partial Discharge source to each sensor is occluded by internal structures of the equipment under test, as this results in a longer path being taken around the obstruction. In this study, Particle Swarm Optimisation has been applied to the problem of Partial Discharge location in order to assess its effectiveness. This search technique simulates the manner in which a flock of birds searches for food, where the 'birds' (particles) are represented by a vector denoting their position in 3D space and the location of the 'food' is the best location found that satisfies the time-differences-of-arrival of the sensors. The search first sought to minimize a ‿blind' least squares function, which was then extended to assign credit to each sensor's contribution to the error associated with each particle's position due to the path between the particles and the sensors being occluded by objects in the test area. This approach produced locations that were competitive with or slightly better than those obtained by another method of Partial Discharge location which estimates the actual path travelled by the signal using a Cartesian model of the environment in which the Partial Discharge occurs.