随着智能电网的发展,部署了大量监控设备形成传感器网络,产生了大量的异构数据,传统数据管理办法难以有效利用海量数据。电力知识图谱具有很强的知识表示和推理能力,可以全面地、条理化地展示电力业务对象的相关业务信息、技术知识、行业标准以及这些信息的内在联系。本文利用图卷积网络构建了电网调度规则的知识图模型,并基于图神经网络提取知识图谱特征。具体来说,首先利用知识提取方法从智能电网调度控制系统中获取规范中的实体词和关系词,明确对象实体集合、消除数据歧义。然后提取实体间的关系构建领域知识图谱,规范化表达电力智能调度的对象关系网络。最后,利用图神经网络模型学习电网调度知识图特征,挖掘实体间隐藏关系,为制定电网调度系统技术提供全面的决策依据。本研究涉及的智能电网调度控制系统技术规范由国家电网东北分公司提供,并在该真实数据集上对本文所提方法进行了验证。

With the development of smart power grid, aSquantitySof monitoring devices are deployed to form sensor networks, resulting in a large amount of heterogeneous data. Traditional data management methods are difficult to effectively utilize the massive data. Because of the strong knowledge representation and reasoning ability, power knowledge graph can comprehensively and systematically display the business information, technical knowledge, industry standards and the internal relationship of these information. In this paper, a knowledge graph model of power grid scheduling rules is constructed by graph convolutional network (GCN), and the features of the knowledge graph are extracted based on graph neural network (GNN).SSpecifically, we obtain the entity words and relation words in the specifications from the smart grid dispatching control system with the knowledge extraction method, so as to clarify the object entity set and eliminate data ambiguity. Then, get the relationship between entities to construct domain knowledge graph, and normalize the object relation network of power intelligent dispatching. Finally, we learn the characteristics of the grid dispatching knowledge graph and mine the hidden relations between entities with the graph neural network, which can provide a comprehensive decision basis for the development of the grid dispatching system technology. The technical specification of smart grid dispatching control system involved in this paper is provided by The Northeast Branch of State Grid, and the method proposed in this paper is verified on the real data set.

TP 391.4

国家自然科学基金面上项目(62076047) ，国家电网有限公司科技项目（52992620003L）