摘要
组合模型能提高电力系统暂态稳定评估的分类性能。文中构建了12组输入特征,在IEEE16节点和IEEE50节点测试系统上生成了24个样本集。提出了一种测试分类综合指标。在24个样本集上比较了单个(神经网络、决策树、K最近邻法和支持向量机)和组合(装袋、提升、堆栈和随机森林)的暂态稳定评估模型测试指标发现,单个评估模型中,K最近邻法分类性能最好;组合方法均能提高分类性能,其中随机森林分类性能最好,其次是堆栈(支持向量机、K最近邻法、决策树)、提升–决策树和装袋–决策树。
Using combined models, the performance of power system transient stability assessment can be improved. In this paper 12 groups of model features are constructed and 24 sample sets are generated by 1EEE 16-bus testing system and IEEE 50-bus testing system. On this basis, a comprehensive index for testing and classification is proposed. By means of the 24 sample sets, the indices from single transient stability assessment models such as artificial neural network, decision tree, K nearest neighbor and support vector machine are compared with those of combined transient stability assessment models such as bagging, boosting, stacking and random forest, and it is found that that within single assessment models the K nearest neighbor possesses the best classification performance; the classification performance can be improved by combined models, and in the order of classification performance, the ordering of combined models is as following： the random forest, the stacking combining with support vector machine, K nearest neighbor and decision tree, the boosting-decision tree and the bagging-decision tree.
出处
《电网技术》
EI
CSCD
北大核心
2008年第23期19-23,共5页
Power System Technology
基金
教育部霍英东青年教师基金(101060)
四川省杰出青年基金(07ZQ026-012)。
关键词
暂态稳定评估
组合模型
装袋
提升
堆栈
随机森林
transient stability assessment
combined models
bagging
boosting
stacking
random forest
