随机动态经济调度问题的Dantzig-Wolfe分解及其并行算法  

Dantzig-wolfe Decomposition and Parallel Algorithm for Stochastic Dynamic Economic Dispatch

作　　者：黄启文 陆文甜 刘明波[1] HUANG Qiwen;LU Wentian;LIU Mingbo(School of Electric Power,South China University of Technology,Guangzhou 510641,Guangdong Province,China)

机构地区：[1]华南理工大学电力学院

出　　处：《电网技术》2019年第12期4398-4405,共8页Power System Technology

基　　金：国家重点基础研究发展计划项目(973项目)(2013CB228205)~~

摘　　要：随着电力系统规模增大和场景法抽样场景数量增加,在求解含风电接入电力系统的随机动态经济调度问题时,计算将遭遇"维数灾"问题而无法求解。因此提出Dantzig-Wolfe分解及其并行优化算法,基于Monte Carlo抽样方法建立了以最小化发电成本为目标的随机动态经济调度模型,将随机优化问题转化为大规模确定性优化问题。采用Dantzig-Wolfe分解对其实施场景解耦,将大规模问题分解为上层主问题和一系列低维度的下层子问题,通过迭代求解主、子问题得到原问题最优解。在迭代过程中,利用改进次梯度法改善了算法收敛性。运用GAMS平台的网格计算工具构建了快速求解误差场景子问题的并行计算框架。既降低了计算机内存需求,从而实现高维问题的求解,又提高了求解速度。以含风电场的IEEE 39节点系统和某省级实际电力系统为例,验证了文中所提方法的正确性与有效性。With growth of power system scale and increase of sampled scenario number, "curse of dimensionality" occurs when solving stochastic dynamic economic dispatch(SDED) problem of power system with random wind power. For this reason, a parallel optimization algorithm based on Dantzig-Wolfe decomposition is proposed. A SDED model with the objective of minimizing power generation cost is established based on Monte Carlo sampling method, so that the stochastic problem is transformed into a large-scale deterministic problem. The large-scale problem is decomposed into a main problem and a set of small-dimension sub-problems with Dantzig-Wolfe decomposition, and optimal solution can be obtained after iteratively solving them. The algorithm convergence is improved by means of subgradient method. A parallel computing framework for quickly solving sub-problems is constructed by leveraging grid computing facility(GCF) on GAMS. This method reduces memory requirement and speeds up the solution. Numerical simulations on IEEE 39-node system and an actual provincial power system with wind farms demonstrate correctness and effectiveness of the proposed method.

关 键 词：场景法 风电场 随机动态经济调度 Dantzig-Wolfe分解 场景解耦 并行计算 

分 类 号：TM721[电气工程—电力系统及自动化]