Knowledge Graph Construction Base on Power Accident Emergency Plan

Power dispatching systems have accumulated a large amount of semi-structured and unstructured accident emergency plan texts in the last decades. How to take full use of these text data for efficiently formulating power accident emergency plans has become a crucial problem. Knowledge graph provides a feasible way to facilitate to acquire emergent accidents and query related plan contents. However, the knowledge graph for power dispatching and emergency plans is a challenging task due to the large amount of semi-structured/unstructured text data. In this paper, in order to improve the efficiency of dealing with power failures and power accident emergency plans, we propose a knowledge graph construction method based on the historical accident emergency plan documents. Firstly, we analyze the document structure of the historical accident emergency plan texts, and further formulate the domain knowledge by designing the top-level ontology of power accident emergency plan. Secondly, we utilize neural network model to extract relations among sentences in each part of contents by analyzing the sentence description characteristics of the power accident emergency plan text. A method of entity relationship extraction is proposed using the DocuNet model for extract relations from the power accident emergency plan text. We made experiments over the historical power accident emergency plan text against some popular methods, and the results show that the entity relation extraction method has good extraction effect and adaptability on the power accident emergency plan text data set.