Solving realistic large-scale ill-conditioned power flow cases based on combination of numerical solvers

Abstract

With the increasing electricity consumption and difficulty in upgrading exis-ting infrastructures, ill-conditioned power flow (PF) cases are becomingmore frequent nowadays. In this context, classical robust solvers may beunsuitable for realistic networks, which typically encompass thousands ofbuses, because of their high computational burden or low convergence rate.This article tackles this issue by proposing a novel PF solver, which presentsacceptable robustness and efficiency in solving large-scale ill-conditioned sys-tems. The proposed algorithm collects the advantage of various numericalsolvers, which by separate present different weaknesses, but actuating in coor-dination their strengths can be jointly exploited. More precisely, the robustForward-Euler and Trapezoidal rules are combined with the efficient Darvishicubic technique. Thereby, an original predictor-corrector algorithm is devel-oped to effectively coordinate the different numerical algorithms involved,obtaining a robust but efficient yet solution procedure. Various large-scale ill-conditioned benchmark systems are studied under different stressing condi-tions. The results obtained with the developed technique are promising, out-performing other robust and standard PF solvers.