Parallel solution of large power system networks using the Multi-Area Thévenin Equivalents (MATE) algorithm

Extending the Multi-Area Thévenin Equivalents method for parallel solutions of bulk power systems

“…In this case, applying (34)-(39) into (31) and (32) Moreover, further neglecting the sequential and communication factors yields the theoretical speedup of the MATE algorithm of p 1+c /2 time over sparsity-oriented solvers. As also shown by Alvarado [41] and verified by Tomim [28], the parameter c tends to zero in modern computer architectures, the theoretical MATE speedup can be estimated in p/2, where p is the number of subsystems. As for the theoretical efficiency of the method, for a large enough number of subsystems, an efficiency of about 50% should be expected.…”

“…The parameter b max , in (27), represents the maximum number of border nodes in the subsystems. Table 1 Partitioning of the WECC system according to multi-level recursive bisection algorithm.…”

“…As summarized by the authors, the MATE algorithm embodies the concepts of multi-node Thévenin equivalents, diakoptics, and the modified nodal analysis by Ho et al [23], in a simple and general formulation. The first dedicated hardware architecture for MATE was presented in [24,25], while the first general-purpose implementation on an off-the-shelf PC cluster, including sparsity techniques, was performed in [26][27][28]. In this last implementation, solutions of the Western Electricity Coordinating Council (WECC) system ($15,000 buses) with the parallel MATE algorithm achieved a 7-time speedup on 15 processors as compared to a single-processor sequential sparsity-oriented solver.…”

Graphics-Processing-Unit-Based Acceleration of Electromagnetic Transients Simulation

“…Using several different test cases, the authors' earlier work [9], showed that GPU-based EMT simulation produces exactly the same result as the commercial tools. Other researchers also used this approach for creating large test cases [5], [12], [13], [30].…”

“…Many attempts have been made to speed up EMT simulation, such as: parallel implementation of the EMT algorithm [4], use of the multi area Thevenin equivalents (MATE) algorithm [5], multi-processor based parallel EMT simulation [6], etc. These approaches have improved the performance of EMT simulations.…”