Extracting oil workover construction information using natural language methods to improve intelligent efficiency
Traditional methods for extracting workover knowledge have the shortcomings of low efficiency provided by manpower and inability to handle large-scale data,resulting in a lack of scientificity in the formulation level of workover measures.For this reason,in the entity extraction stage,an improved attention mechanism based on label weights is designed,which,together with the pre-trained weight model and Bidirectional Long Short-Term Memory(BiLSTM),forms a workover knowledge entity extraction model.In the association rule mining stage,an improved Apriori algorithm that integrates the Bayesian method and the Hash tree is proposed,thus forming a two-stage intelligent analysis and mining method for workover knowledge oriented to the texts of construction plans.Upon being applied in Dagang Oilfield,The results indicate that the recognition accuracy of the workover knowledge entity extraction model can reach 81.83%.The number of frequent itemsets mined by the improved Apriori model is 814,with 515 strongly associated entity combinations,and the computational efficiency of the association rules is increased by 34.38%.The intelligent analysis and mining method for workover knowledge proposed in this article can enhance the efficiency of workover knowledge extraction,providing ideas for data extraction and digital construction in the field of petroleum engineering.
Artificial intelligenceBig dataAlgorithmWorkoverDigital economyLaboratory testNew quality productivityOil and gas reform
万方数据
维普
