This paper proposes a fill-in dropping strategy during ILU preconditioner construction making use of two criteria; one to eliminate the L factor fill-ins based on the resulting error and the other based on a relative tolerance for U fill-ins. Normally, the fill-in dropping rule is based on levels of fills, numerical values, amount, or a combination of the former two. This combination rule is very well-documented in the literature but for solving the load flow sublinear problem this conventional rule may introduce large errors due to inadequate droppings, increasing the number of operations and the CPU time as well, turning poor the iterative method performance. It is shown the influence of the reordering scheme over that scenario and numerical experiments typify the problem and corroborate the better quality of the proposed rule.
Preconditioners; Incomplete Factorization; Krylov subspace iterative methods; Algebraic Linear Equations; Load-Flow
