Named Entity Recognition for Smart Grid Operation and Inspection Domain using Attention Mechanism
Lizong Zhang, Ende Hu, Xianke Zhou, Tianheng Chen, Saisai Yang, Yuxiao Zhang · 2022
With the rapid development of smart power grid systems, the field of power grid operation and maintenance is in urgent need of intelligent construction. In the daily operation of the power system, a large amount of operation and maintenance texts are accumulated. Using these data for named entity recognition to support the construction of knowledge graphs in the power grid domain and finally realize the decision intelligence in the power grid operation and inspection domain is the key of this paper. In this paper, we propose a named entity recognition algorithm based on$\text{BERT}+\text{BiLSTM}+\text{CRF}$, which combines attention mechanism and word-character joint embedding vector, for the text features in the field of grid operation and inspection. Experiments show that the algorithm proposed in this paper achieves 91.47%, 88.88% and 90.16% in accuracy, recall and$F_{1}$values while the training convergence speed is smaller than that of the baseline model, which is 1.27% 4.06% better than the traditional model. In particular, the improvement in position of fault and type of fault entity recognition is obvious, with 0.25% 6.79% and 3.28% 9.21% respectively, which verifies the feasibility and effectiveness of the algorithm model proposed in this paper and thus effectively supports the subsequent construction of the knowledge graph and intelligent system in the field of power grid operation and maintenance.