A Comparative Study of Particle Filter Resampling Techniques for Power Systems State Estimation

Abstract

The effect of the resampling stage on the overall performance of the particle filter state estimation for power systems is investigated. As each general particle filter algorithm includes three steps of the prediction, update and resampling, the resampling step is essential to avoid the sample degeneracy problem. In this paper, the standard particle filter is applied to the problem of the decentralized state estimation in power systems. Various existing resampling techniques are described, tested, and evaluated in different power systems with different fault scenarios. The resampling methods are compared based on their mean square error, their needed computation time and their performance against various data quality issues such as delays, missing measurements, etc. More efficient resampling methods are then determined for three cases of the power system cases: a single-machine-infinite-bus system, the medium size two-area four-machine system and the large-scale New-England system considering different fault scenarios.