【摘要】 提出了一种改进的多中心-校正内点法,该方法采用超立体空间映射技术,通过合理配置一些关键映射参数改善映射空间的数值结构,使算法在每次迭代时能获得较大的迭代步长和较好的中心方向,从而加快算法的收敛。通过IEEE14、30、57、118节点4个测试系统的仿真计算表明,该算法收敛快,其迭代次数与每次迭代进行4次中心-校正计算的多中心-校正内点法基本相当,而且鲁棒性好,未出现数值稳定问题。文章还对算法收敛判据的设置和初值的选取作了较详细的讨论。更多还原

【Abstract】 An improved multiple centrality-correction interior point method is proposed. Using hypercube maping technique and rationally collocating some key maping parameters to improve the numerical structure of maping space, a longer iteration step length and a better central path can be obtained in each iteration by the proposed algorithm, therefore the convergence is speeded up. The simulation results from IEEE 14-, 30-, 57- and 118-node system show that the proposed algorithm possesses fast convergence speed, and its iteration times is basically equivalent to that of the multiple centrality-correction interior point method which carries out four times of centrality-correction calculation in each iteration. The proposed method has good robustness and the numerical stability problem has not appeared. The setting of convergence criteria and the selection of initial value are also researched in detail.更多还原