Current security model simulations are based on component stability limits that are determined very infrequently, using worst-case models. Improved stability limits based on the actual state of the network would greatly improve the accuracy of these simulations. Stability limits are based on transient stability analysis, which examines the dynamic behavior of the network following a power flow disturbance. This involves the solution of a set of coupled differential-algebraic equations, which is a highly compute intensive problem. We are developing parallel algorithms to solve these equations, and the associated sparse matrix equations that are required for power systems problems.
Analysis of the output of the transient stability program is a technical art, requiring trained, experienced engineers. In order for the results to be available on-line in near real-time, we are planning to use an expert system to identify results as being transient stable or unstable. We are currently developing a knowledge base to be used for the expert system.
Another compute-intensive application being studied is the problem of unit commitment. This is a large-scale optimization problem, concerned with efficiently matching the available generation capacity to the expected load. We are also investigating other possible applications of HPC, from databases and billing to the simulation of nuclear power stations.
An ordering of a sparse matrix, representing the connectivity of an
electrical power grid,
into a block-diagonal-bordered form that is more suitable for
parallelization. Blue denotes non-zeros, and red is fillin (elements
that were originally zero but become non-zero during factorization).