SCALE: A hybrid MPI and multithreading based work stealing approach for massive contingency analysis in power systems

• Parallel contingency analysis.
• High performance computing (HPC).
• Dynamic load balancing.
• Work-stealing scheduling.
• Master–slave scheduling.

In this paper, we present SCALE, a hybrid message passing interface (MPI) and multithreading based work Stealing approach for massive Contingency AnaLysis in powEr systems. SCALE performs time domain simulation of power systems using efficient numerical algorithms and scales the approach for analyzing a large number of contingencies using MPI. For achieving dynamic load balancing, SCALE uses efficient implementation of the work stealing algorithm. SCALE uses a hybrid MPI and multithreading based implementation, where MPI is used for communication between different nodes (processors) and multithreading is used within each processor to facilitate implementation of non-blocking version of work-stealing algorithm. We have evaluated SCALE on a large, 13,029 bus system using thousands of contingencies. Also, we have compared the dynamic load balancing based scheduling approach of SCALE with a state-of-art scheduling approach, namely master–slave scheduling approach. The results show that SCALE is effective in analyzing a large number of contingencies and scales well to a large number of processors.