Architecture and Implementation of Electricity Statistics Knowledge Graphs

Abstract

With the continuous advancement of the digital power business, the data and knowledge in the field of power show explosive growth. To better manage the scattered data and exploit its potential value, an electricity statistics knowledge graph is structured, in which the data is extracted into triplet. And the electricity knowledge graph is stored by the index matrix and the adjacency matrix for the sake of fast search. On the basis of the constructed graph, four applications have been developed, namely search function, perspective generation, perspective recommendation and electricity forecasting. Search function provides the basis for information acquisition. Perspective is statistical analysis of data flow based on knowledge graph. Perspective recommendation and electricity forecasting are advanced applications of knowledge graph which is combined with artificial intelligence. Perspective recommendation use graph convolution neural network to recommend noteworthy perspectives based on historical operations. Electricity forecast is a bottom-up forecast based on the integration of knowledge graph and RNN. The knowledge graph constructed based on the data of a municipal power grid shows that the constructed knowledge graph has faster search speed and more functions.