MASTER: A JAVA Based Work-Stealing Technique For Parallel Contingency Analysis
Siddhartha Kumar Khaitan · HAL (Le Centre pour la Communication Scientifique Directe) · 2016
In this paper, we present MASTER, a Java based multithreaded work-stealing technique for parallel contingency analysis in power systems. MASTER analyzes contingencies using time domain simulation and scales contingency analysis task to multiple cores using multithreading in Java. To achieve load balancing, MASTER uses efficient implementation of work-stealing algorithm. We discuss several implementation issues and design time choices which are crucial for achieving efficient implementation. Experiments performed with contingencies of a 13029 bus power system shows that MASTER provides high computational gains and provides much better load-balancing than the conventional scheduling techniques.