中文摘要：

提出一种基于聚类算法的电网安全评估新思路。以关键稳态状态量为输入，应用聚类算法提取样本空间分布知识，利用所获知识实现系统稳定水平评估。聚类算法以样本为起点构造子空间，不断扩展子空间以获得包含数据分布结构的最优子空间，最优子空间的聚合构成聚类结果，并以类边界样本展示训练集空间分布结构。算法对数据形状适应性强，适合增量式数据集的挖掘。在IEEE两个测试系统上的应用结果证实所提电网安全评估思路的有效性。

英文摘要：

A new method for power system security assessment based on the unsupervised clustering algorithm is proposed in the paper. Using the clustering algorithm extract the knowledge of spatial distribution of the sample which is used to achieve system stability assessment with the steady state variables as inputs. Clustering algorithm to construct the subspace as a starting point of sample, and extend the subspace in order to obtain the optimal subspace which contain the data distribution structure, and the aggregation of optimal subspace constitute the clustering results, and use the boundary samples to show the spatial distribution structure of the training set. The approach is adaptive to the different data shapes and incremental data set conveniently, application results on the two IEEE testing systems can confirm the effectiveness of new method.