引用本文: | 方斯顿,程浩忠,徐国栋,曾平良,姚良忠.基于Nataf变换和准蒙特卡洛模拟的随机潮流方法[J].电力自动化设备,2015,35(8): |
| FANG Sidun,CHENG Haozhong,XU Guodong,ZENG Pingliang,YAO Liangzhong.Probabilistic load flow method based on Nataf transformation and quasi Monte Carlo simulation[J].Electric Power Automation Equipment,2015,35(8): |
|
摘要: |
提出一种可处理输入变量相关性的随机潮流方法。该方法基于Nataf变换和准蒙特卡洛模拟,使用奇异值分解处理对称非正定的相关系数矩阵。对IEEE 30节点系统和某实际大区域电网的仿真证明了所提方法的有效性和普适性。仿真结果表明:与传统基于Cholesky分解的排序方法相比,奇异值分解可在不增加计算代价的同时灵活处理非正定的相关系数矩阵;而相比于普通基于拉丁超立方的方法,所提方法收敛更快,相同样本规模下的计算精度更高,特别是输出变量标准差的精度。 |
关键词: 随机潮流 Nataf变换 准蒙特卡洛模拟 奇异值分解 |
DOI: |
分类号: |
基金项目:国家自然科学基金重点资助项目(51337005);国家重点基础研究发展计划(973计划)资助项目(2014CB23903) |
|
Probabilistic load flow method based on Nataf transformation and quasi Monte Carlo simulation |
FANG Sidun1, CHENG Haozhong1, XU Guodong1, ZENG Pingliang2, YAO Liangzhong2
|
1.Key Laboratory of Control of Power Transmission and Conversion,Ministry of Education,
Shanghai Jiao Tong University,Shanghai 200240,China;2.China Electric Power Research Institute,Beijing 100192,China
|
Abstract: |
A probabilistic load flow method capable of processing the correlation among input variables is proposed based on the Nataf transformation and quasi Monte Carlo simulation,which adopts the singular value decomposition to deal with the symmetric non-positive correlation coefficient matrix. The effectiveness and universality of the proposed method are verified by the simulations for IEEE 30 -bus system and a practical large power grid. The simulative results show that,compared with the ranking method based on Cholesky decomposition,the singular value decomposition can flexibly deal with the non-positive correlation coefficient matrix without additional computational burden,while,compared with the normal method based on Latin hypercube under the same sample size,the proposed method has higher convergence speed and higher computational accuracy,especially the standard deviation accuracy of output variables. |
Key words: probabilistic load flow Nataf transformation quasi Monte Carlo simulation singular value decomposition |