Date of Award

May 2019

Degree Type

Thesis

Degree Name

Master of Science

Department

Engineering

First Advisor

Lingfeng Wang

Committee Members

Lingfeng Wang, David Yu, Guangwu Xu

Keywords

Diagnosis, Fault, Machine Learning, MMC-HVDC

Abstract

With the development of the HVDC system, MMC-HVDC is now the most advanced technology that has been put into use. In power systems, faults happen during the operation due to natural reasons or devices physical issues, which would cause serious economic losses and other implications. Thus, fault detection and analysis are extremely important, especially in the HVDC system. Existing works in literature mainly focus on the faults detection and analysis on the system side such as short circuit of the AC side, and open circuit of the DC side. However, little attention has been paid to the fault detection and analysis inside the converters. With the technology development of converter devices, replacing the whole converter becomes more expensive. Thus, my research mainly focuses on the detection and classification of the faults within the internal of the MMC module.

In this research, an SPS model of MMC-HVDC is built as the example. Faults including short circuit and open circuit located inside the MMC module are simulated. Machine learning algorithms are chosen as the tool to achieve the goal of detecting faults and locating the fault position inside the MMC module precisely. After comparing the basic characteristics and properly application situations of various methods of machine learning, Coarse KNN, Complex Tree and Bagged Tree (Random Forest) are deployed to solve the problem. The performance of the methods are analyzed and compared, to get the most proper method in solving the problem.

Available for download on Wednesday, November 11, 2020

Share

COinS