摘要
特征选择和空间降维是电力系统暂态稳定评估中的关键步骤。针对国内外现有文献所提方法中存在的效率低、分解子集操作结果不理想等问题,提出了基于极限学习机和遗传算法的输入特征选择方法。首先运用遗传算法实现特征选择,再输入优选后的子集,利用极限学习机构造分类器加以稳定性评判。其中,适应度函数考虑了两个要素:一是所选特征子集应对分类结果起到较为重要的作用;二是用作输入的特征项尽可能精简。在英格兰10机39节点系统中进行仿真计算,结果表明,进行特征选择后分类效果优于未进行特征选择情况,与其他文献所选的特征子集相比,该方法所选特征子集的分类准确率更高,证明了其有效性和优越性。
Feature selection and input dimension reduction are important for the transient stability assessment of power system. To solve the problems in the existing feature selection methods, such as low efficiency and unsatisfactory de- composing subset result, a method is proposed based on extreme learning machine （ELM） and genetic algorithm. First, genetic algorithm is used to realize feature selection. Then the selected feature is input into ELM classifier for transient stability assessment. There are two factors in constructing the fitness function ： one is that the selected feature subset should have a greater contribution to the classification ; the other is that the adopted input features should be as less as possible. The application to a 10-Machine 39-Bus New England power system indicates that the effect is obvious- ly better after feature selection. Compared with other methods in the literature, the classification accuracy of the pro- posed approach is higher, which demonstrates its validity and advantage.
作者
卢锦玲
於慧敏
LU Jinling YU Huimin(School of Electrical and Electronic Engineering, North China Electric Power University, Baoding 071003, China)
出处
《电力系统及其自动化学报》
CSCD
北大核心
2016年第12期103-108,共6页
Proceedings of the CSU-EPSA
关键词
电力系统
暂态稳定评估
特征选择
遗传算法
极限学习机
power system
transient stability assessment
feature selection
genetic algorithm
extreme learning ma-chine（ELM）
作者简介
卢锦玲（197l-），女，博士，副教授，研究方向为电力系统运行、分析与控制。Email：lujinling@126．com
於慧敏（1992-），女，硕士研究生，研究方向为电力系统运行、分析与控制。Email：yuhuimin_huadian@163．com
