摘要
利用支持向量机(SVM)方法进行暂态稳定判别时,输入特征的选择是影响最终结果的最重要因素。传统启发式和试探式方法不能从根本上解决输入特征选择的问题。本文利用信息融合思想,在构造的具有不同输入特征的多组子分类器的基础上,对子分类器的结果在输出空间再进行信息融合,以提高分类准确率。文中从不同角度启发式的构造了 4,构成四组弱分类器。以这四组弱分类器为子分类器,再构造一个融合 SVM 对几种子分类器的结果以回归方式进行融合,作为最终判别结果。IEEE 39-BUS 和IEEE145-BUS 测试系统上进行的仿真表明,弱分类器的分类性能经过融合得到明显强化,融合后的结果比任何一种子分类器的结果以及一次包含所有输入特征的结果都更准确。该方法为在线快速进行暂态稳定计算提供了一条重要途径。
In the assessment of the transient stability with the Support Vector Machine (SVM), the choices of the input features are the most important factors to the final results. The traditional heuristic methods and the tentative methods cannot solve this problems radically. This paper, with the idea of information fusion, proposes a two-layer SVM classifier model to improve the precision of the classification. In this paper, four different groups of input features are built based on the heuristic knowledge of different angles to form four weak classifiers. By taking these four weak classifiers as the sub-classifiers, a fusional SVM is built to fuse the results of the sub-classifiers in a manner of regression, whose result is taken as the final judgment result. Simulation results on the IEEE 39-BUS and IEEE 145BUS test system show that the classification performance of the weak classifiers is evidently strengthened after fusion, and that the result after the fusion is more precise than any of the single results and the result including all the input features at once. The proposed method is a promising tool of fast computation for on-line transient stability assessment.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2005年第6期17-23,共7页
Proceedings of the CSEE
基金
西北电网公司项目(04-dd02)。
