基于交叠概率的暂态电压稳定评估特征选择

An Overlapping Probability Based Feature Selection Method for Evaluation of Transient Voltage Stability

针对电力系统的高维特征量,提出了一种能有效降低维数的特征选择方法。该方法以最小概率落入类别间的不可判别区域为原则选取特征组合。算法核心在于特征组合过滤判据的确定,判据的形成基于样本点落入类别间交叠区域的概率,方法简易直观。针对系统的暂态电压稳定评估问题,首先构建了一组暂态电压稳定评估的原始特征属性集,将经特征选择降维后的特征组合作为决策树的输入,并用10倍交叉验证方法对评估结果进行验证。2个标准系统的算例表明,通过该方法对暂态电压稳定评估进行特征选择得到的特征属性组合在电压稳定评估的应用上具有更高的准确率。

In allusion to high-dimensional feature quantities of power system, a feature selection method that can effectively reduce the dimensionality is proposed. In the proposed method, according to the principle of the minimum probability of test samples falling into indiscriminative zone between the categories the feature combinations are selected. The core of the proposed algorithm consists in the determination of filtration criterion of feature combinations, and the criterion is formed based on the probability by which the sample points fall into the overlapping zone between categories, so the proposed method is simple and intuitional. To assess transient voltage stability of power system, firstly an attribute set of original features for the assessment of transient voltage stability is established; then the feature combinations after the dimensionality reduction by feature selection are taken as the input of the decision tree and by use of 10-fold cross-validation the assessment results are validated. Calculation results of New England 39-bus network and IEEE 145-bus system show that the stability is more accurate by combinations obtained by the voltage stability assessment, assessment of transient voltage applying the feature attribute proposed method to transient