基于混合高斯模型的相关非高斯输入变量随机潮流计算  

Probabilistic load flow with non-Gaussian correlated input variables based on Gaussian mixture model

作　　者：黄煜[1] 徐青山[1] 刘建坤 卫鹏 Huang Yu Xu Qingshan Liu Jiankun Wei Peng(School of Electrical Engineering, Southeast University, Nanjing 210096, China Jiangsu Electric Power Research Institute, Nanjing 210003, China)

机构地区：[1]东南大学电气工程学院,南京210096 [2]江苏省电力公司电力科学研究院,南京210003

出　　处：《东南大学学报（自然科学版）》2017年第2期291-298,共8页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(51577028);国家电网公司科技资助项目

摘　　要：提出一种考虑输入变量相关性的随机潮流计算方法.该方法针对系统中的非高斯输入变量,建立其混合高斯模型(GMM).在此基础上,引入高斯分量组合算法(GCCM),通过多次加权最小二乘计算(WLS)直接求得输出变量的概率分布.研究限制GMM中高斯分量个数的约简方法,以减少WLS运算次数.对IEEE-30节点系统的仿真和误差分析表明,GMM具有拟合精度高、适用性广的特点.所提方法与MCS的计算结果基本一致,但计算效率有了显著提高,并且算法的速度和精度与WLS运算次数有关.An algorithm for probabilistic load flowconsidering the correlation between input variables was proposed. A Gaussian mixture model（ GMM） was established by the algorithm to represent non-Gaussian input variables in the system. On such a basis,a Gaussian component combination method（ GCCM） was introduced and the marginal distribution of any output variable was directly obtained from multiple weighted least square runs（ WLS）. A study was also carried out to reduce the number of trials by limiting the number of Gaussian components. The simulation and error analysis on IEEE-30 test system indicated that GMMhad the features of high fitting precision and wide applicability. The results obtained from the proposed method are identical to that of MCS and the computational efficiency is obviously improved. The effectiveness and the accuracy are proved to be closely related to operation times of WLS.

关 键 词：混合高斯模型 约简算法 相关性 高斯分量组合 随机潮流 

分 类 号：TM74[电气工程—电力系统及自动化]