摘要
常规的勘探开发数据自动分类以语义理解为主,模态较为单一,影响最终的分类准确性。设计了基于多模态信息融合的油田综合勘探开发数据自动分类方法,将数据特征分为结构化、非结构化、计划、运行、成果等类型,以三维分析的形式提取数据文本特征。基于多模态信息融合构建开发数据自动分类模型,将文本、图像等数据拼接成新的数据格式,融合不同模态的数据特征,确保数据分类的准确性。剔除模态不平衡冗余数据,解决文本、图像、传感器等不同模态数据的不平衡问题,进一步满足数据自动分类需求。采用对比实验,验证了该方法的自动分类准确性更高,能够应用于实际生产中。
Conventional automatic classification of exploration and development data is mainly based on semantic understanding,and the mode is relatively single,which affects the final classification accuracy.An automatic classification method for oilfield comprehensive exploration and development data based on multimodal information integration is designed,the characteristics of data is classified into structured,unstructured,planned,operational,and achievement data.The text features of data are extracted in the form of three-dimensional analysis.An automatic classification model for development data based on multimodal information integration is constructed,data such as text and images are concatenated into a new data format,data features from different modalities are integrated to ensure the accuracy of data classification.Imbalanced and redundant data are eliminated,to solve the problem of imbalance of data in different modalities such as text,images,and sensors,and to further meet the requirements of automatic data classification.Through comparative experiments,it has been verified that the method has higher accuracy in automatic classification,and can be applied in practical production.
作者
龙喜彬
安精文
郭鑫
杨迎春
杨春琴
Long Xibin;An Jingwen;Guo Xin;Yang Yingchun;Yang Chunqin(Exploration and Development Research Institute of Northwest Oilfield Branch of China Petroleum&Chemical Corporation,Urumqi,830011,China)
出处
《石油化工自动化》
2025年第1期73-76,共4页
Automation in Petro-chemical Industry
关键词
多模态信息融合
油田
综合勘探
开发数据
自动分类方法
multimodal information integration
oilfield
comprehensive exploration
develop data
automatic classification method
