Evaluation of two algorithms of multiple corrections on complementarity conditions

In this paper, a careful analysis and evaluation for two approaches based on multiple corrections on complementarity conditions in the nonlinear predictor-corrector primal-dual interior point algorithm (PCPDIPA) is carried out. Unlike the other methodologies that improve the performance of PCPDIPA from the perspective of problem modeling/formulation, these two approaches directly tackle the centrality issue of the algorithm by re-using the available matrix factorization. Several parameters controlling the performance of these two approaches are identified by a series of numerical testing on several large-scale cases. Numerical comparison results are provided as indices for choosing appropriate centrality correction methodology for other power system applications.