Two-Stage Fault Classification Algorithm for Real Fault Data in Transmission Lines

Abstract

Fault classification in power transmission lines is important in distance relaying for identifying the accurate phases implicated in the fault occurrence. Generally, the accuracy of fault classification algorithms is evaluated by simulation data, which shows quite different characteristics from real fault data. Also, most of the previous works on fault classification used a single-stage method such as a rule-based algorithm or machine learning-based algorithm. Because of the diverse characteristics of real fault data, the performance of the single-stage method is limited. To address these issues, this paper proposes a novel two-stage algorithm that combines the strengths of rule-based and machine-learning algorithms to improve the accuracy of real fault data. A case study using real fault data shows that the proposed two-stage algorithm outperforms other conventional single-stage algorithms.

Keywords

• Two-stage algorithm
• rule-based
• artificial neural network
• root mean square