Research On Reservoir Characteristics Extraction And Knowledge Graph Construction From Geological Document

With the increasing development of oil and gas exploration theories and technologies in China,all kinds of geological data describing the geological environment,accumulation factors,oil-bearing horizons and other related characteristics of oil and gas reservoirs are accumulating rapidly,and the volume is increasing.It brings challenges to data management and information services.There are many kinds of data in the field of oil and gas,most of which are unstructured data,such as survey reports,production records and research documents,mainly stored in the form of documents,which have few characteristics for query and lack of knowledge description and association.Only by keyword matching and coarse-grained topic similarity search,it is not possible to quickly and accurately obtain information from complex data,and it is difficult to meet the needs of knowledge query for massive geological text,resulting in low data utilization rate,and a large amount of knowledge contained in unstructured text cannot be fully and effectively displayed.From the development of information technology and the trend of digital oil field construction in China,oil and gas geology data service is changing from data service to knowledge service.The advent of knowledge graph technology has provided an opportunity for the knowledge-based service of oil and gas geology data.It is possible to provide users with efficient knowledge services by automatically extracting information contained in documents and forming structured knowledge.This paper focuses on the characteristics of oil and gas reservoirs in geological texts,combined with the characteristics of oil and gas geological data,reorganizes the knowledge segments in the data and forms knowledge elements.The current data retrieval mode is enriched by extracting the characteristics,attributes and relationships of oil and gas reservoirs in the knowledge element.The knowledge map of hydrocarbon accumulation is gradually constructed by bottom-up method and applied to the knowledge retrieval service of oil and gas geological data.Through the functions of literature subject selection and knowledge association query,users can improve the efficiency of data information acquisition and assist them to discover implicit associations contained in data,which will help to improve the accuracy of petroleum geology research and the reliability of oil and gas exploration decision-making.The main works of this paper are as follows:(1)Knowledge element extraction from oil and gas geological texts based on hierarchical topic.Based on the analysis of the structure and subject level of oil and gas geological data,the extraction of knowledge elements in geological texts is studied.The text topic analysis technology is used to obtain the topic features of text paragraphs.Taking into account the strong hierarchy and clustering of texts in the field of oil and gas,we study the use of professional vocabulary in the field to constrain the hierarchical topic model,so as to improve the effect of topic extraction.At the same time,the description of knowledge elements of oil and gas reservoir characteristics is studied in combination with the structure of oil and gas geologic data and practical application requirements.In this paper,a topic knowledge element extraction method based on local feature algorithm joint strategy and dynamic programming strategy is proposed.The similarity of graph headings and paragraphs,matching method of chart and pointer are proposed to extract chart knowledge elements.(2)Reservoir feature extraction based on domain knowledge and machine learningAccounting the correspondence and relevance of the reservoir knowledge contained in the oil and gas geological data,the paper extracts reservoir feature information from the geological document,combining with prior knowledge such as the oil and gas domain ontology,professional thesaurus,and relational database metadata.For the extraction of overlapping entities and overlapping relationships in oil and gas reservoir characteristics,this paper studies the advantages of domain ontology in the semantic description of concepts and relationships,puts forward a relationship extraction model based on vocabulary feature enhancement and improved label strategy,and achieves the extraction of feature information such as entities,attributes and relationships of oil and gas reservoir features in text knowledge elements.(3)Construction of oil and gas accumulation knowledge graph based on multifeature correlationAiming at the deficiencies in knowledge representation and correlation of existing models,the relationship between various characteristics of reservoirs is studied.This paper presents a topic-based extension of knowledge representation and vectorization method for ternary reservoirs to solve the problem of knowledge representation for oil and gas reservoirs.A knowledge representation and quantitative method based on topic extension is proposed to solve the problem of knowledge representation in oil and gas reservoirs.Next,the multi-feature association method of knowledge elements in highdimensional space is studied to solve the abstract representation and association of knowledge elements in oil and gas reservoirs.Finally,the knowledge graph of oil and gas accumulation is applied to the knowledge service system,and the knowledge retrieval service and knowledge element graphic correlation query are provided,which verifies the adaptability and practicability of the proposed method.