Support Vector Machine Based Fault Location Identification in Microgrids Using Interharmonic Injection

Energies

Mehdi Baharizadeh¹  Alireza Forouzesh²  MohammadS. Golsorkhi²  Mehdi Savaghebi³ 

[1] Department of Electrical Engineering, Khomeinishahr Branch, Islamic Azad University, Isfahan 8418148499, Iran;Department of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan 8415683111, Iran;Electrical Engineering Section, Department of Mechanical and Electrical Engineering, University of Southern Denmark, Campusvej 55, 5230 Odense, Denmark;

关键词: fault location;    harmonics;    machine learning;    microgrid;    power electronics;    protection;   

DOI  :  10.3390/en14082317

来源: DOAJ

【 摘 要 】

This paper proposes an algorithm for detection and identification of the location of short circuit faults in islanded AC microgrids (MGs) with meshed topology. Considering the low level of fault current and dependency of the current angle on the control strategies, the legacy overcurrent protection schemes are not effective in in islanded MGs. To overcome this issue, the proposed algorithm detects faults based on the rms voltages of the distributed energy resources (DERs) by means of support vector machine classifiers. Upon detection of a fault, the DER which is electrically closest to the fault injects three interharmonic currents. The faulty zone is identified by comparing the magnitude of the interharmonic currents flowing through each zone. Then, the second DER connected to the faulty zone injects distinctive interharmonic currents and the resulting interharmonic voltages are measured at the terminal of each of these DERs. Using the interharmonic voltages as its features, a multi-class support vector machine identifies the fault location within the faulty zone. Simulations are conducted on a test MG to obtain a dataset comprising scenarios with different fault locations, varying fault impedances, and changing loads. The test results show that the proposed algorithm reliably detects the faults and the precision of fault location identification is above 90%.

【 授权许可】

Unknown