基于神经网络的暂态稳定评估输入空间可分性

Input Space Separability Analysis of Artificial Neural Networks Based Transient Stability Assessment

介绍一种数据结构分析算法，并将其应用于基于神经网络的暂态稳定评估输入空间可分性分析。使用该算法，可将稳定评估的高维输入空间近似地转化为一个二维空间，从而可直接观察其可分性并评估特征的表达能力。在两个标准电力系统中的应用结果表明了该数据结构分析算法的有效性。同时，该算法也可用于电力系统中的其它模式分类问题。

This paper describes a data-structure analysis algorithm and applies it to separability analysis of input spaces for transient stability assessment. Using this analysis algorithm, an input space with a high dimension for transient stability assessment can be transformed into a two-dimensional space, therefore, its separability can be directly observed and the representing capability of the selected attributes can be assessed. Numerical results in two well-known power systems demonstrate the validity of the data structure analysis algorithm. Additionally, this algorithm can also be used to other pattern classification problems in power systems.