Feature Selection for Fast and Reliable Passive Islanding Detection in Synchronous Generation

  • Abdullah Sawas

Student thesis: Master's Thesis

Abstract

The hype of smart grids promises to give solutions for the issues associated with conventional central electricity generation systems. Since in smart grids a consumer may become a producer for electricity, new decision techniques for controlling the ?ow of electricity are needed. Taking advantage of the advancement in machine learning approaches and digital communication tools, smart grids would make intelligent decisions in order to efficiently supply power to consumers. One of the challenges associated with smart grids is the islanding issue where part of the grid with a generator is isolated from the main utility. Detecting this state is of great importance to utility companies because of a number of issues related to the safety, stability and quality of power delivered to customers. To avoid these problems, utilities need to quickly identify the operating conditions and act upon by either switching the operation into islanded mode or simply disconnect the distributed generator in the island. This thesis investigates the prominent parameters that can be employed in the detection algorithm under various operating scenarios. Using data mining approaches, a feature selection method is utilized to rank the features in a way that increases the detection method's reliability and speed. We found that some features are not very sensitive to the change in detection time while others have been affected intensively. In addition, we studied the effect of load types on the selection of features and observed that some prominent features in one type of load are still important in the aggregated loads. Finally, we proposed a multi-stages detection method that uses a set of features within different time windows which increased the information available and allowed for early detection of some cases.
Date of Award2011
Original languageAmerican English
SupervisorWei Lee Woon (Supervisor)

Keywords

  • Islanding
  • Smart Grids
  • Electricity Generation Systems

Cite this

'