An Electric System Abnormal Analysis Framework based on Natural Language Processing

The stable operation of power grid system is essential for our daily lives. The on-site patrol and examining play an important rule to detect the potential damage and stabilize the power grid system. In this paper, we propose a framework to analyze possible errors in electric system by natural language processing method so that the on-site workers can directly describe the phenomenon and get the instruction to deal with the situation. The framework applies the word2vec algorithm to calculate the similarity between the words, and Earth Movers' Distance method is adopted to compute the overall semantic similarity between the on-site situation and the history record. Based on the similar records, the next step operation is generated to guide the on-site workers. We implement the proposed framework for electric distribution network and the experimental results indicate that the proposed method has better accuracy (93%) compared with existing methods (84%) and can be used to improve the management of power grid system.