ElGENANALYSlS OF LARGE ELECTRIC POWER SYSTEMS

Modern electric power systems are large and complicated, and, in many regions, the generation and transmission systems are operating near their limits. Eigenanalysis is one of the tools used to analyze the behavior of these systems. Standard eigenvalue methods require that simplified models be used for these analyses; however, these simplified models do not adequately model all of the characteristics of large power systems. Thus, new eigenanalysis methods that can analyze detailed power system models are required. The primary objectives of the work described in this report were I) to determine the availability of eigenanalysis algorithms that are better than methods currently being applied and that could be used an large power systems and 2) to determine if vector supercomputers could be used to significantly increase the size of power systems that can be analyzed by a standard power system eigenanalysis code. At the request of the Bonneville Power Administration, the Pacific Northwest Laboratory (PNL) conducted a literature review of methods currently used for the eigenanalysis of large electric power systems, as well as of general eigenanalysis algorithms that are applicable to large power systems. PNL found that a number of methods are currently being used for the this purpose, and all seem to work fairly well. Furthermore, most of the general eigenanalysis techniques that are applicable to power systems have been tried on these systems, and most seem to work fairly well. One of these techniques, a variation of the Arnoldi method, has been incorporated into a standard power system eigenanalysis package. Overall, it appears that the general purpose eigenanalysis methods are more versatile than most of the other methods that have been used for power systems eigenanalysis. In addition, they are generally easier to use. For some problems, however, it appears that some of the other eigenanalysis methods may be better. Power systems eigenanalysis requires the computation of eigenvalues of nonsymmetric matrices. Such computations are fairly difficult, however, and they constitute an area of active research. Thus, research in this area should be closely monitored, and, as new methods become available, they should be tested on large power systems. PNL also investigated the use of vector supercomputers to enlarge the size of power systems that can be analyzed by MASS, a standard power system eigenanalysis code. MASS was converted to run on a Cray supercomputer. On a conventional computer, MASS is limited to power system problems with about 500 states because of computer time constraints. Running MASS on a Cray X-MP EA/232, however, PNL found that a problem with about 2200 states could be solved in about 26 minutes. Furthermore, by moving to a larger Cray, it should be possible to use MASS to analyze power systems with 5,000 to 10,000 states. Problems with 5,000 states would probably take about 5 Cray hours, while a problem with 10,000 states would probably take about 43 Cray hours, though the actual execution times will depend on the type of Cray used. Problems requiring 5 Cray hours are not uncommon. Problems requiring 43 Cray hours, however, would be fairly expensive. Thus, power systems with about 5000 states probably represent an upper limit on the size of problems that one would want to routinely solve using a Cray version of MASS