Solution of large power-system networks by ordered elimination: a comparison of ordering schemes

A variety of different ordering schemes are used with sparsity-programmed Gaussian elimination to solve a range of practical power-system networks, with a view to assessing their relative merits for general-purpose applications. It is shown that, for networks of the sizes and types studied, the choice between the more successful schemes is dependent mainly on the relative amounts of work required to establish the orderings. One simple and popular scheme, involving pivot selection at each stage of the elimination process on a row-sparsity basis, emerges as the most attractive.