提高电力系统潮流计算收敛性方法的比较分析

Comparative Analysis Among Prevailing Methods for Enhancing Convergence of Power Flow Calculation

为提高大型电力系统潮流计算的收敛性,现阶段主要采用最优乘子法、张量法和自适应LevenbergMarquardt(LM)方法。分别对这3种方法在数值特点、稀疏实现和计算量方面进行了分析和比较。分析发现,最优乘子法计算永不发散且对牛顿法的修改和增加计算量最少,但计算结果容易陷入局部解;基于插值的张量法对重负荷系统补偿效果较好,但数值稳定性有待改进;自适应LM法能够得到潮流方程的精确最小二乘解,但迭代步的计算较为复杂。采用1个标准系统和2个实际系统进行仿真实验,仿真计算结果表明,相比其他两种方法,自适应LM法具有更好的鲁棒性和数值稳定性,但其实用化依赖于高效的稀疏实现。

In order to improve the convergence of power flow calculation in large-scale power system, currently the optimal multiplier method, tensor method and self-adaptive Levenberg-Marquardt （LM） method are widely utilized. This paper analyzes and compares these three methods in terms of numerical property, sparse implementation and equiva- lent computational efforts. Through the analysis of each method, it can be asserted that the optimal multiplier method never diverges and requires the least extra computational efforts compared to Newton method, but results are liable to be trapped into local solutions. The interpolation tensor method has good compensation effect for systems with heavy load, whereas its numerical stability needs to be boosted. The self-adaptive LM method can obtain the exact least square solution of power flow equations, although each iteration step is intricate. Simulation results from 1 standard IEEE test system and 2 practical power systems indicate that the self-adaptive LM method is much more robust and nu- merically stable than other methods, and its practical application relies on an efficient sparse implementation.