摘要
Abstract
After more than 60 years of exploration and development,Daqing Oilfield is facing increasing challenges of theoretical innovation,technological progress and field implementation.In this process,traditional mathematical models approached to the bottleneck in processing complex geological data and engineering data,requiring new re-search paradigms.In recent years,large models,especially multimodal generative large models,as a kind of intelli-gent technology based on big data and large models,can handle large-scale,multi-source and heterogeneous data,and are expected to better deal with the complexity and uncertainty in oil and gas exploration and development,re-shape digital process and become a new development opportunity for oil and gas industry.Starting from technical re-quirements and era background,technical system of oil and gas large model is explored,with envisage for the con-struction of oil and gas large models.Specific scheme starts from construction goals and principles,designs techni-cal architecture and provides implementation path.It plans 4 stages horizontally and 4 research fields vertically,in-cluding designed 15 key tasks,and further plans the construction of 4 supporting systems including data base,knowledge base,research achievements database and collaborative research and formulates 13 key technologies for large model implementation,gradually promoting the understanding of"data","graphs"and"geobodies"by large models.Finally,application scenarios of large models in future oil and gas research fields are prospected.关键词
油气大模型/多模态/生成式预训练模型/垂直领域模型/指令微调Key words
oil and gas large model/multimodal/generative pre-trained model/vertical domain model/instruc-tion fine-tuning分类
能源科技
