Security Assessment of Power System with Artificial Neural Network

Student: Thomas Wei Kwang Lee

Supervisor: Zhao Dong

Category: Engineering Thesis Project - Power

Modern Power System is of great complexity due to the constantly changing load conditions. In addition, geographically isolated generators require long transmission lines, posing high chances of disturbances and contingencies. In order to operate the Power System economically to remain competitive, operating status of the Power System has to be defined. This process is known as Security Assessment.

The aim of this thesis is to propose a framework using Neural Network based techniques for real-time power system Security Assessment to meet the needs of current power industry in a deregulated environment. The techniques proposed provides options for independent system operators of a competitive electricity market in obtaining the system Security information with sufficient details of considerations in real time to meet their liabilities for Power System Security. Power System is a highly interconnected nonlinear dynamic system which makes it extremely difficult to model and assess the Security considerations in real time. Often, Off-line Security Assessment has to be carried out to form lookup tables for later on line usage; or simplifications have to be made to enable fast Security Assessment. The proposed framework captures the Power System dynamics with the learning capability of Neural Networks and provides fast real time response.

In addition, ANN learning rules, architecture and construction algorithm will be critically analysed. Two test systems are designed; namely, a 3-Bus Test System for the analysis of Q-Index and a 6-Bus Test System for the analysis of Transmission Line flow. Details on the Methodology of the building, training, testing and implementation of an ANN on Q-Index analysis and Transmission Line Flow based on the two designed Power System will also be discussed.

Thesis Document (PDF)