基于CNN-LSTM的油田产量预测模型  

Oilfield Production Forecasting Model Based on CNN-LSTM

作　　者：张汉文 辛显康 喻高明[1,2,3] ZHANG Hanwen;XIN Xiankang;YU Gaoming(School of Petroleum Engineering,Yangtze University,Wuhan 430100,China;Hubei Key Laboratory of Drilling and Production Engineering for Oil and Gas,Yangtze University,Wuhan 430100,China;National Engineering Research Center for Oil and Gas Drilling and Completion Technology,Yangtze University,Wuhan 430100,China)

机构地区：[1]长江大学石油工程学院,湖北武汉430100 [2]长江大学油气钻采工程湖北省重点实验室,湖北武汉430100 [3]长江大学油气钻完井技术国家工程研究中心,湖北武汉430100

出　　处：《现代信息科技》2025年第5期33-38,共6页Modern Information Technology

基　　金：国家自然科学基金青年科学基金(52104020)。

摘　　要：油田产量预测是制定油田生产开发方案的基础工作。随着数据量的不断增加,传统方法面临诸多挑战。通过机器学习建立产量预测模型,已成为一种更加准确高效的方法。文章将卷积神经网络(CNN)与长短期记忆网络(LSTM)两种基础模型相结合,构建了CNN-LSTM模型。通过CNN提取油田数据中某些突出的特征,并将这些特征作为构建LSTM时间序列数据集的输入,揭示了数据中隐藏的特征与空间相互作用关系,进而提高预测结果的精确度。此外,实验结果表明,相比单一的LSTM模型和CNN模型,CNN-LSTM模型在预测性能上具有显著优势,能够更准确地预测油田生产数据,为油田的后续开发提供可靠的数据支持。Oilfield production forecasting is a fundamental task for formulating production and development scheme.With the increasing volume of data,traditional methods face significant challenges.Establishing forecasting models through Machine Learning has become a more accurate and efficient method.This paper combines two fundamental models,Convolutional Neural Networks(CNN)and Long Short-Term Memory(LSTM),to construct a CNN-LSTM model.It extracts some prominent features from the oilfield data by CNN,and uses the features as input to construct the LSTM time series dataset.It reveals the hidden relationship between features and spatial interaction in data,and then enhances the accuracy of the forecasting results.In addition,the experiment results demonstrate that compared with standalone LSTM model and CNN model,the CNN-LSTM model has significant advantages in terms of forecasting performance.It can forecast the oilfield production data more accurately and provide reliable data support for the subsequent oilfield development.

关 键 词：CNN-LSTM 机器学习 产量预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TE348[自动化与计算机技术—控制科学与工程]