Improving an interior-point-based OPF by dynamic adjustments of step sizes and tolerances

This paper presents an efficient interior point algorithm for optimal power flow (OPF) problems, in particular, the real power dispatch and the reactive power dispatch problems. The nonlinear OPF problem is solved by a predictor-corrector primal-dual log-barrier (PCPDLB) method as a sequence of linearized sub-problems. Besides discussing the problem formulation, the paper offers a detailed description of the PCPDLB algorithm; it also addresses several implementation issues such as the determination of barrier parameter and the customization of initial points for OPF problems. In addition, practical issues on how to choose linear step sizes and convergence criteria are investigated to evaluate their impact on the performance of the algorithm. Some heuristics of dynamically adjusting step sizes and tolerance are proposed which significantly improve OPF solution speed. Computational results on power systems of 118 and 1062 buses are presented and discussed. Comparisons with other variants of primal-dual log-barrier methods are also provided to demonstrate the superiority of the proposed predictor-corrector interior point algorithm.