A hybrid intelligent busbar protection strategy using hyperbolic S‐transforms and extreme learning machines. Issue 12 (9th July 2021)

• Record Type: Journal Article
• Authors: Rahimnejad, Abolfazl; Gil, Milad; Abdoos, Aliakbar; Gadsden, S. Andrew
• Is Part Of: Engineering reports. Volume 3:Issue 12(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-07-09
• Subjects: CT saturation -- differential protection -- extreme learning machine (ELM) -- hyperbolic S‐transform -- inrush current -- intelligent busbar protection -- support vector machine (SVM); Engineering -- Periodicals; Computer science -- Periodicals; 620.005
• DOI: 10.1002/eng2.12438
• Languages: English
• ISSNs: 2577-8196

Abstract

Abstract: In power systems, busbars connect important components such as generators, transmission lines, and loads. A typical fault occurrence on the busbar may result in the isolation of faulty sections from other normally operating parts of the system resulting from differential protection operation. Although the main busbars' protection scheme is differential protection, its operation is significantly affected by magnetic saturation of the current transformer (CT), particularly during external fault occurrence or energizing power transformers. Saturation of the CT may generate a spurious differential current and is the main reason for the differential scheme malfunctioning. Previous research presented different methods to modify and improve busbars' differential protection scheme. However, there has been lack of a comprehensive study to assess the efficiency of the busbar protection scheme regarding all involved, and influencing aspects including various fault types, energizing power transformer, (high) fault resistance, fault angle (changing from 0° to 360°), and the angle of the sources. Thus, in this study, a hybrid intelligent busbar protection scheme is proposed and the effects of all these factors are investigated. The proposed strategy utilizes the hyperbolic S‐transform as a signal processing technique to extract an efficient feature that is able to discriminate internal faults from other abnormal modes, that is, external faults and inrush current under CT saturation. To obtain this goal, a learning‐based classification method known as extreme learning machines is used to classify the system conditions based on the selected features. The proposed protection scheme was found to have low sensitivity to CT saturation and noise and was able to accurately detect internal faults from half a cycle to one cycle of the power system depending on the fault resistance. Abstract : 3D representation of the extracted features used in a new hybrid intelligent busbar protection strategy based on hyperbolic S‐transforms and extreme learning machines. The features allow for accurate discrimination of internal faults from external faults and inrush currents in busbar systems.