基于流形排序的电网截面数据检索

The Grid Section Data Retrieval Based on Manifold Ranking

随着电力系统信息化的发展,电网数据具有数量大、类型多、维度高的特点。针对在数据检索时多维度查询效率不高,检索结果无法多维度整体匹配的问题,提出一种基于流形排序的电网截面数据检索方法;该方法将电网截面数据描述成多维向量空间中的对应点,创建加权图模型。通过考虑数据的整体近似流形结构来获得检索结果,使之与源查询之间具有较高的相关性;使用置信传播分配排序分数,提高了检索结果的准确性,有效避免相关性度量对高维数据查询处理的不足。

With the development of power system information,grid data has the characteristics of large quantities,various types and high dimensions. Aiming at problems of multidimensional query with low efficiencies and the retrieval results without multidimensional integral matching in data retrieval,a power grid cross section data retrieval method based on manifold sort was put forward,which describes power grid cross section data as multi-dimensional corresponding points in the vector space,constructing a weighted graph model and using the global manifold structure within the data to obtain the correlation among the queries. Using belief propagation distribution to sort points can improve the accuracy of the retrieval results and effectively avoid the lack of correlation measure of highdimensional data query processing.