A new numeric busbar protection scheme using Bayes point machine

Abstract

Schemes adopted for protection of busbar face major challenges in terms of speed, sensitivity, and immunity against current transformer (CT) saturation. This paper reports a new busbar protection scheme using Bayes Point Machine (BPM). A substation model with double bus configuration has been simulated in PSCAD and one cycle post-fault current samples are utilized to identify the fault zone. The proposed algorithm has been tested for diversified fault scenarios including the cases for which the conventional line differential protection scheme (87L) is likely to maloperate. BPM is found to be more than 99% accurate while identifying the correct fault zone. A comparative evaluation suggests the superiority of BPM among other machine learning based classifiers.