A Differential Particle Swarm Optimization-based Support Vector Machine Classifier for Fault Diagnosis in Power Distribution Systems

作者：Cho Ming Yuan; Hoang Thi Thom^*

来源：Advances in Electrical and Computer Engineering, 2017, 17(3): 51-60.

DOI：10.4316/AECE.2017.03007

• fault diagnosis
• particle swarm optimization
• power distribution lines
• reflectometry
• support vector machines

This paper proposes a new differential particle swarm optimization (DPSO) method for obtaining optimum support vector machine (SVM) parameters used for electrical fault diagnosis in radial distribution systems. Further, a multiple-stage DPSO-SVM classifier is developed to enhance classification accuracy in the fault diagnosis. Also, time-domain reflectometry (TDR) method with pseudo-random binary sequence (PRBS) excitation is utilized for generating the dataset required for validating this proposed approach. According to the characteristic of echo responses found in different types of faults, 12 features are extracted as input vectors for purposes of classification. The proposed fault diagnosis approach is tested on a typical radial distribution system to classify ten types of short-circuit faults accurately. Further, to demonstrate the superiority of the proposed DPSO algorithm, comparative studies of fault diagnosis are performed using SVM having parameters selected using cross-validation, GA and PSO. The overall classification accuracy obtained for fault diagnosis is 98.5%, which shows the effectiveness of the proposed approach.

• 出版日期2017