Artificial intelligence based approach for secured protection during stressed conditions

摘要

Maloperation of distance relays due to stressed conditions have caused a number of blackouts in the history of Power System protection. Limitations of the present blocking schemes have encouraged computational intelligent techniques to be used for assisting the conventional distance relay. For this purpose two of the most popular techniques have been employed namely Support Vector Machine (SVM) and Decision Tree (DT). SVM being a black-box model has limitations in interpreting the results obtained. While DT is a transparent method that not only makes the process comprehensible but also provides with the dominant features participating in the technique. A two level classifier scheme is built using both the models SVM and DT for distinguishing fault from stressed conditions and further differentiating the stressed conditions to be either Power Swing or Voltage Instability. It has been tested on WSCC 9 bus system with Dig SILENT Power Factory as the simulation software.