Design And Implementation Of Knowledge Extraction System In Knowledge Graph Of Electric Power Dispatch

With the rapid development of computer technology and the combined use of artificial intelligence and multiple fields,the convenience provided by artificial intelligence-related technologies in production and life has also made people realize the various potentials of artificial intelligence technology.Among them,the knowledge map involves with the development of recent years,the various technologies have become a research hotspot.In the field of electric power,electric power dispatching is an effective management method and management method used to ensure that various types of electric power production work are carried out in an orderly manner,to ensure that the internal power grid can ensure its safe and stable operation,and to provide reliable power supply externally.In order to improve the efficiency of power dispatching,combined with the development of current knowledge graph technology,a knowledge extraction system in the field of power dispatching is designed and implemented,and the system is used as one of the preliminary work in the construction of knowledge graphs in the subsequent power dispatching field to ensure subsequent knowledge The rest of the work in the map construction can proceed smoothly.In order to solve the problem of knowledge extraction in the knowledge map in the field of power dispatching,because there is no relevant knowledge extraction technology in the field of power dispatching that can be used directly,so research and refer to some research materials and usage of knowledge extraction technology in other fields,and then combine the undergraduate questions The actual situation.Some techniques used in the design and implementation of the knowledge extraction system in the field of power dispatching are used.In data preprocessing,some data cleaning techniques are used to process the repeated values and noise data in the data;in the entity extraction part,the Transformer model is used to obtain the positional embedding of the sentence instead of Word2 Vec to obtain the word vector of the sentence,and the BiGRU model is used to pass positional embedding to get the label sequence of the sentence,after using the CRF to operate on the label sequence,the original sentence corresponding to the entity and the attribute of the sentence are obtained;the Attention model is used in the relationship extraction in the BiGRU model.The label sequence and the entity extraction result are both In the case of partial integration,the relationship between entities is extracted;in the data storage part,Redis is used to store and persist the data;in the model implementation part,the programming technology combining Python language and TensorFlow is used.In the training part of the model,the entity extraction part uses two methods to train the model in the system.These two training methods are mainly different training methods for the three models of Transformer model,BiGRU model and CRF model.The first one is Transformer The model is not trained and uses the Google-trained model,but only the BiGRU and CRF models are trained together;the second is to train the three models together.The data storage part builds a Redis database storage system,stores the data in the Redis database,and stores the extracted data persistently after storage to ensure that the data can continue to be used again after the program exits.In the process of designing and implementing the knowledge extraction system in the knowledge map of power dispatching,the specific features of knowledge extraction in this field and the research data extracted from knowledge extraction in other fields are referenced to implement and test the relevant functions required by this topic.The accuracy rate of entity extraction in this system reaches 70%,which meets the expected goal of functional design;the F1 value of relationship extraction reaches close to 70%,indicating the feasibility of relationship extraction general energy;the feasibility of data storage ensures that the system has the function of data storage Implementation.Through the realization of each function of each module in the system,the function of knowledge extraction in the knowledge graph in the field of power dispatching required by this system is finally realized.It provides part of the preliminary work for the construction of knowledge graphs in the power field,and realizes the automatic extraction of knowledge extraction in the early stage of the construction of knowledge graphs in the power field,which reduces costs and improves work efficiency.