Abstract: Defects of the relay protection devices occur from time to time, threatening its reliability as the first line of defense, thereby affecting the safe and stable operation of the electric power grid. With the expansion of the power grid and the increase in the number of the repay protection devices, the operation and maintenance require proper intelligent assistance to reduce the work pressure. In the meantime, the accumulation of the historical defect data of the protection devices contains a lot of information which may be useful for the data mining of the intelligent assistance. In view of this, this paper proposes an entity relation extraction for the construction of the relay protection device defect knowledge graph based on the actual relay protection device defect data in a certain area and the professional dictionaries. Firstly, based on the traditional ontology construction idea and the TF-IDF(term frequency - inverse document frequency) method, the terminology in the relay protection field is extracted and the concepts and their relationships are defined so as to realize the construction of the ontology for the relay protection device defect. Secondly, according to the text features of the relay protection device defect records, an entity relationship extraction based on the grammar rules is proposed, which realizes the adjacent lexeme entity relationship extraction. Thirdly, an entity relationship extraction based on the remote supervised learning is proposed, which achieves the entity relationship extraction from the long-spaced lexeme. Further, combined with the relationship extraction evaluation, the entity relationship extraction with the comprehensive grammar rules and the remote supervision is realized. Finally, the examples of the ontology and the relationship extraction for the defect of the relay protection device are partially visualized using the Protégé.