Towards a computationally feasible on-line voltage instability index

Several researchers have suggested characterising a power system voltage collapse using the singularity of the Jacobian of the network descripter load-flow equations. It has been proposed to use the minimum singular value of the Jacobian as a security index for monitoring how closely a power system is operating to a voltage instability. However, a singular value decomposition (SVD) computation is an iterative process with a number of arithmetic operations proportional to the cubic power of the matrix size. State-of-the-art SVD computation schemes using large arrays of parallel processors that improve the computation speed linearly and quadratically are discussed. A comparison is made between the computation speed and the number of arithmetic units required for various parallel architecture schemes as well as with the traditional serial computer. >