牛顿法潮流计算中两种稀疏存储方式的效率研究

Study on the Efficiency of Two Sparse Storage Modes for Newton Power Flow Calculation

研究了数组存储和链表存储这2种稀疏技术在牛顿法潮流计算中的效率问题,分析了它们在内存开销上的差别。内存分析和不同规模系统的算例表明:链表存储比数组存储多用的内存基本上是系统支路数、节点数的线性函数;采用列主元高斯消去法解修正方程时,链表存储相对于数组存储,以付出较小的内存空间赢得了计算时间的很大节省,系统规模较大时,链表存储较数组存储具有明显的优势;系统规模越大,优势越明显。在采用LDU分解时,链表存储较数组存储要稍快。

In this paper, the efficiency problems of two sparse techniques, namely, storage of array and storage of linking list, in Newton power flow calculation have been studied. The differences of EMS memory overheads of the two modes have been also analyzed. EMS memory analysis and calculation for different--scale systems indicate that： the reason for the more memory needed by storage of linking list than that of storage of array is that the linear function of the number of buses and branches; when the modified linear equations are solved with Gauss Elimination Method with column pivots, the storage of linking list occupies a little more memory, but saves much more time than that of array; the larger storage the system scale, the superior the linking list compared with the storage of array; the calculation speed of the storage of linking list is only a little faster than that of the storage of array while the modified linear equations are solved with LDU factorization.