融合邻域粗糙约简与深度森林的电力系统暂态稳定评估  

Power System Transient Stability Assessment Based on Hybrid Neighborhood Rough Reduction and Deep Forest

作　　者：李兵洋 肖健梅[1] 王锡淮[1] Li Bingyang;Xiao Jianmei;Wang Xihuai(Department of Electrical Automation Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海海事大学电气自动化系,上海201306

出　　处：《电工技术学报》2020年第15期3245-3257,共13页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(61573240);上海海事大学博士研究生创新基金(2017ycx084)资助项目。

摘　　要：在实际电网运行中,暂态稳定样本与暂态失稳样本间呈现出明显的不平衡关系,且误判失稳样本与误判稳定样本的代价不同。当前基于数据挖掘的暂态稳定评估方法多基于浅层模型,对误判暂态失稳样本的重视不够,且评估精度有待进一步提高。基于此,提出一种融合邻域粗糙约简与深度森林的电力系统暂态稳定评估方法。利用邻域粗糙集在不同粒度级别下寻找多组不同的最优特征子集以对原始特征进行再表征,通过深度森林的级联结构实现对原始暂态特征的表征学习,强化特征量与暂态稳定状态间的非线性映射关系;引入加权投票机制,提高分类过程对暂态失稳样本的重视。在IEEE 10机39节点系统上的实验结果表明,所提方法能够在提升评估精度的同时,有效降低了对暂态失稳样本的误判,在不同数据规模和不同程度的不平衡样本数据上均具有较好的表现,具有一定的鲁棒性和适用性。In the operational process of power grid, transient stable samples and transient unstable samples are obviously imbalanced, and the cost of misclassifying stable samples is unequal to that of unstable samples. The existing transient stability assessment methods using data mining techniques are mostly based on shallow models, which pay little attention to the situation of misclassifying transient unstable samples. Moreover, the evaluation accuracy needs to be further improved. This paper proposes a power system transient stability assessment method integrating neighborhood rough reduction and deep forest. By using neighborhood rough sets at different granularity levels, several optimal feature subsets can be obtained to re-represent the original feature space. The cascade structure of deep forest can further strength the representation learning ability,which can reinforce the nonlinear mapping relation between features and transient stability state. The employment of weighted voting mechanism can make the learning process pay more attention to transient unstable samples. The experimental results on IEEE 10 machine 39 bus system show that the proposed method can effectively improve the evaluation accuracy and reduce the misclassification rate of transient unstable samples. Moreover, it also has a good performance on data sets with different scale and imbalance degrees, which is robust and applicable.

关 键 词：邻域粗糙集 暂态稳定评估 深度森林 电力系统 

分 类 号：TM712[电气工程—电力系统及自动化]