针对传统聚类算方法无法应对新能源大规模接入配电网造成数据多元且高维的问题，提出了基于多元数据的配电网典型场景生成方法，并定义了配电网长短期安全运行评价指标体系。文章采用主成分分析法对配电网高维数据降维，并建立基于DBSCAN（density-based spatial clustering of applications with noise）和改进K均值的两阶段聚类模型，以此得到保留极端场景的典型场景集。同时根据配电网安全运行要素，定义配电网长短期安全运行指标体系。以改进的IEEE 33节点配电网为例进行算例分析，对比多种聚类方法，验证两阶段聚类方法兼具有效性和高效性，所得的典型场景应用于电压越限薄弱点识别，具有高效性和精确性。
英文摘要:
In response to the problem of data diversity and high dimensionality caused by the large-scale integration of new energy into the distribution network, traditional clustering algorithms are unable to cope. A typical scenario generation method for distribution network based on multi-source data is proposed, and an evaluation index system for long short-term safe operation of distribution network is defined. Principal component analysis is used to reduce the dimensionality of high-dimensional data in the distribution network, and a two-stage clustering model based on density-based spatial clustering of applications with noise(DBSCAN) and improved K-means is established to obtain a typical scenario set that preserves extreme scenarios. Based on the safety operation elements of the distribution network, it defines the long short-term safety operation index system of the distribution network. An improved IEEE 33-node distribution network is simulated and compared with various clustering algorithms to verify the effectiveness and efficiency of the two-stage clustering algorithm. The typical scenarios obtained are applied to identify the weak points of voltage violation, which has high efficiency and accuracy.