Skip to Main Content
This paper presents a parallel direct linear solver based on a message-passing distributed-memory multiprocessor architecture such as a cluster of workstations. The results show that the new algorithm can achieve nearly linear speedup for two large-scale power system cases on a small cluster of GNU/Linux dual-processor workstations. The workstations are connected via 100 Mbit/s Ethernet, i.e., the parallel machine consists of hardware readily found in any engineering department. Based on the presented parallel direct linear solver, it is possible to parallelize totally the Newton power flow solution process. In addition, the METIS-based partitioning scheme can handle common control devices such as PV-PQ switching. Furthermore, by tuning the vertex and branch weights, the performance of the power flow solution can be optimized for the available hardware. For a workstation cluster on 100 Mbit/s Ethernet, the speedup appears to saturate beyond eight processors due to load imbalance and the aggregate growth of the partition separators. Nevertheless, the message-passing distributed-memory multiprocessor architecture can be used in other power system applications, such as state estimation and transient stability. Furthermore, an iterative linear solver could improve scalability.