基于决策表-粗糙集理论的动态安全分析神经网络输入特征优选

提出了一种动态安全分析神经网络输入特征的优选方法。该方法将决策表最优特征子集理论与粗糙集理论的元素离散化方法相结合,在保证较高精度的运行点分类和稳定裕度计算的基础上,从大维数的注入元中优选特征元作为神经网络输入特征,以降低其输入特征的个数,较好地解决了神经网络动态安全域方法用于大系统所面临的“维灾难”问题。通过对电力系统分析综合程序(PSASP)中EPRI-36测试系统的仿真计算,验证了该方法的有效性。

ANN INJECTION FEATURES OPTIMIZATION OF DYNAMIC SECURITY ANALYSIS BASED ON DECISION TABLES-ROUGH SET THEORY

An ANN injection features optimization method of dynamic security analysis is presented. It combines optimal feature subset of decision tables with element discretization of rough-set. Based on the relatively high accuracy of running point' s classification and stability margins computation, the feature elements are selected from the lager dimensions injections and regarded as ANN injection features. So the number of the injection features can be reduced, and the "dimensions misfortune" problem caused by application of ANN dynamic security regions method to bulk power system is solved. The validity of the method is proved by simulation on PSASP EPRI-36 test system.