基于长短期记忆神经网络的油田新井产油量预测方法
|
摘要
针对油田常用人工智能产油量预测方法无法考虑数据在时间上相关性的问题,提出了采用基于长短期记忆(简称LSTM)神经网络的油田新井产油量预测方法。在分别介绍反向传播(简称BP)神经网络、循环神经网络(简称RNN)、LSTM神经网络原理以及建模步骤的基础上,以某油田新井单井年产油量预测为例,对影响新井单井年产油量的开发指标进行了筛选,对相应LSTM神经网络进行了训练,并对新井单井年产油量进行了预测。将预测结果与支持向量回归模型和BP神经网络进行了对比,结果表明,该预测模型拟合效果更好,预测精度更高。基于LSTM神经网络的预测方法可以作为一种新的人工智能方法用于油田新井产油量的预测,为准确预测油田新井产量,指导油田开发决策提供了一种新的方法。
New well production forecast method based on long-term and short-term memory(LSTM)neural network was proposed to solve the problems that artificial intelligence production prediction method commonly used in oilfields cannot consider the temporal correlation of data with time.Based on the introduction of principle and modeling steps of back propagation(BP)neural network,recurrent neural network(RNN),and LSTM neural network,development indicators affecting yearly oil production of new single well were selected taking the yearly production forecast of new single well of an oilfield as an example,the corresponding LSTM neural network were trained,and the yearly oil production of new single well was forecasted.The forecasted results were compared to those of support vector regression model and BP neural network.The results show that the forecast model has good fitting result with higher forecast accuracy.The forecast method based on LSTM neural network can be used as a new artificial intelligence method for the oil production forecast of new well in oilfields.It is a new method to accurately forecast the oil production of new wells in oilfield and to guide oilfield development decision making.
New well production forecast method based on long-term and short-term memory(LSTM)neural network was proposed to solve the problems that artificial intelligence production prediction method commonly used in oilfields cannot consider the temporal correlation of data with time.Based on the introduction of principle and modeling steps of back propagation(BP)neural network,recurrent neural network(RNN),and LSTM neural network,development indicators affecting yearly oil production of new single well were selected taking the yearly production forecast of new single well of an oilfield as an example,the corresponding LSTM neural network were trained,and the yearly oil production of new single well was forecasted.The forecasted results were compared to those of support vector regression model and BP neural network.The results show that the forecast model has good fitting result with higher forecast accuracy.The forecast method based on LSTM neural network can be used as a new artificial intelligence method for the oil production forecast of new well in oilfields.It is a new method to accurately forecast the oil production of new wells in oilfield and to guide oilfield development decision making.
引文
[1]孙政顺,董范.油田产量模型的非线性回归方法拟和[R].合肥:全球智能控制与自动化大会,2000.SUN Zhengshun,DONG Fan.Nonlinear regression method fitting of oilfield production model[R].Hefei:Global Conference on Intelligent Control and Automation,2000.
[2]王美石,陈祥光,李宇峰.用于油田产量预测的多元线性回归和自回归模型[J].石油规划设计,2005,16(3):19-21.WANG Meishi,CHEN Xiangguang,LI Yufeng.Multivariate linear regression and automatic regression models applied to output prediction of oilfield[J].Petroleum Planning&Engineering,2005,16(3):19-21.
[3]林旺,范洪富,王志平,等.致密油藏体积压裂水平井产量预测研究[J].油气地质与采收率,2018,25(6):107-113.LIN Wang,FAN Hongfu,WANG Zhiping,et al.Prediction on production of volume fractured horizontal well in tight reservoirs[J].Petroleum Geology and Recovery Efficiency,2018,25(6):107-113.
[4]陈民锋,郎兆新.基于自适应遗传算法的油田产量灰色预测模型[J].系统工程学报,2003,18(6):541-546.CHEN Minfeng,LANG Zhaoxin.Production forecast gray model of oilfield based on adaptive genetic algorithm[J].Journal of Systems Engineering,2003,18(6):541-546.
[5]王一淋,赵磊,梅海燕,等.GM(1,1)与支持向量机组合模型在特高含水期油田产量预测中的应用[J].石油天然气学报,2009,31(1):271-273.WANG Yilin,ZHAO Lei,MEI Haiyan,et al.Application of combination model of GM(1,1)and support vector machine in oilfield production prediction in extra high water cut period[J].Journal of Oil and Gas Technology,2009,31(1):271-273.
[6]李志强,赵金洲,胡永全,等.致密油层多区体积压裂产能预测[J].油气地质与采收率,2016,23(1):134-138.LI Zhiqiang,ZHAO Jinzhou,HU Yongquan,et al.Productivity forecast of tight oil reservoirs after multi-zone stimulated reservoir volume fracturing[J].Petroleum Geology and Recovery Efficiency,2016,23(1):134-138.
[7]马福军,李楠,季世侠.新立油田压裂增产量预测模型的建立及应用[J].特种油气藏,2017,24(3):86-89,108.MA Fujun,LI Nan,JI Shixia.Establishment and application of an incremental production prediction model by fracturing in Xinli Oilfield[J].Special Oil&Gas Reservoirs,2017,24(3):86-89,108.
[8]康海亮,林畅松,李洪辉,等.库车坳陷依南地区阿合组致密砂岩气储层特征与有利区带预测[J].石油实验地质,2016,38(2):162-169.KANG Hailiang,LIN Changsong,LI Honghui,et al.Reservoir characteristics and favorable zone prediction of tight sandstone gas of the Ahe Formation in Yinan area,Kuqa Depression[J].Petroleum Geology&Experiment,2016,38(2):162-169.
[9]杨向同,滕起,张杨,等.地质工程一体化支撑下的裂缝性致密砂岩气藏压后评估及产能预测方法研究[J].中国石油勘探,2018,23(2):104-116.YANG Xiangtong,TENG Qi,ZHANG Yang,et al.Post hydraulic fracturing evaluation and productivity prediction method of fractured tight sandstone gas reservoirs supported by geology-engineering integration[J].China Petroleum Exploration,2018,23(2):104-116.
[10]朱小梅,杨先凤,张群燕.LS-SVM-GA算法在油田产量预测中的应用研究[J].煤炭技术,2010,29(11):197-198.ZHU Xiaomei,YANG Xiangfeng,ZHANG Qunyan.Application of LS-SVM-GA algorithm in oil production forecasting[J].Coal Techology,2010,29(11):197-198.
[11]吴新根,葛家理.应用人工神经网络预测油田产量[J].石油勘探与开发,1994,21(3):75-78,131.WU Xingen,GE Jiali.The application of artificial neural network in predicting output of oil fields[J].Petroleum Exploration and Devlopment,1994,21(3):75-78,131.
[12]纪天亮,卢双舫,唐明明,等.致密油藏水平井压裂后产能预测方法[J].大庆石油地质与开发,2016,35(2):165-169.JI Tianliang,LU Shuangfang,TANG Mingming,et al.Productivity predicting method of the fractured horizontal wells in the tight oil reservoirs[J].Petroleum Geology&Oilfield Development in Daqing,2016,35(2):165-169.
[13]李克文,王义龙,苏玉亮.一种基于动态径向基函数神经网络的油田产量预测方法:CN104732303A[P].2015-06-24.LI Kewen,WANG Yilong,SU Yuliang.Oilfield production prediction method based on dynamic radial basis function neural network:CN104732303A[P].2015-06-24.
[14]RUMELHART D E,HINTON G E,WILLIAMS R J.Learning representations by back-propagating errors[J].Nature,1986,323:533-536.
[15]HOCHREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997,9(8):1 735-1 780.
[16]ECK D,SCHMIDHUBER J.Finding temporal structure in music:blues improvisation with LSTM recurrent networks[C].Martigny:Proceedings of the 12th IEEE workshop on neural networks for signal processing,2002.
[17]PéREZ-ORTIZ J A,GERS F A,Eck D,et al.Kalman filters improve LSTM network performance in problems unsolvable by traditional recurrent nets[J].Neural Networks,2003,16(2):241-250.
[18]范竣翔,李琦,朱亚杰,等.基于RNN的空气污染时空预报模型研究[J].测绘科学,2017,42(7):76-83.FAN Junxiang,LI Qi,ZHU Yajie,et al.A spatio-temporal prediction framework for air pollution based on deep RNN[J].Science of Surveying and Mapping,2017,42(7):76-83.
[2]王美石,陈祥光,李宇峰.用于油田产量预测的多元线性回归和自回归模型[J].石油规划设计,2005,16(3):19-21.WANG Meishi,CHEN Xiangguang,LI Yufeng.Multivariate linear regression and automatic regression models applied to output prediction of oilfield[J].Petroleum Planning&Engineering,2005,16(3):19-21.
[3]林旺,范洪富,王志平,等.致密油藏体积压裂水平井产量预测研究[J].油气地质与采收率,2018,25(6):107-113.LIN Wang,FAN Hongfu,WANG Zhiping,et al.Prediction on production of volume fractured horizontal well in tight reservoirs[J].Petroleum Geology and Recovery Efficiency,2018,25(6):107-113.
[4]陈民锋,郎兆新.基于自适应遗传算法的油田产量灰色预测模型[J].系统工程学报,2003,18(6):541-546.CHEN Minfeng,LANG Zhaoxin.Production forecast gray model of oilfield based on adaptive genetic algorithm[J].Journal of Systems Engineering,2003,18(6):541-546.
[5]王一淋,赵磊,梅海燕,等.GM(1,1)与支持向量机组合模型在特高含水期油田产量预测中的应用[J].石油天然气学报,2009,31(1):271-273.WANG Yilin,ZHAO Lei,MEI Haiyan,et al.Application of combination model of GM(1,1)and support vector machine in oilfield production prediction in extra high water cut period[J].Journal of Oil and Gas Technology,2009,31(1):271-273.
[6]李志强,赵金洲,胡永全,等.致密油层多区体积压裂产能预测[J].油气地质与采收率,2016,23(1):134-138.LI Zhiqiang,ZHAO Jinzhou,HU Yongquan,et al.Productivity forecast of tight oil reservoirs after multi-zone stimulated reservoir volume fracturing[J].Petroleum Geology and Recovery Efficiency,2016,23(1):134-138.
[7]马福军,李楠,季世侠.新立油田压裂增产量预测模型的建立及应用[J].特种油气藏,2017,24(3):86-89,108.MA Fujun,LI Nan,JI Shixia.Establishment and application of an incremental production prediction model by fracturing in Xinli Oilfield[J].Special Oil&Gas Reservoirs,2017,24(3):86-89,108.
[8]康海亮,林畅松,李洪辉,等.库车坳陷依南地区阿合组致密砂岩气储层特征与有利区带预测[J].石油实验地质,2016,38(2):162-169.KANG Hailiang,LIN Changsong,LI Honghui,et al.Reservoir characteristics and favorable zone prediction of tight sandstone gas of the Ahe Formation in Yinan area,Kuqa Depression[J].Petroleum Geology&Experiment,2016,38(2):162-169.
[9]杨向同,滕起,张杨,等.地质工程一体化支撑下的裂缝性致密砂岩气藏压后评估及产能预测方法研究[J].中国石油勘探,2018,23(2):104-116.YANG Xiangtong,TENG Qi,ZHANG Yang,et al.Post hydraulic fracturing evaluation and productivity prediction method of fractured tight sandstone gas reservoirs supported by geology-engineering integration[J].China Petroleum Exploration,2018,23(2):104-116.
[10]朱小梅,杨先凤,张群燕.LS-SVM-GA算法在油田产量预测中的应用研究[J].煤炭技术,2010,29(11):197-198.ZHU Xiaomei,YANG Xiangfeng,ZHANG Qunyan.Application of LS-SVM-GA algorithm in oil production forecasting[J].Coal Techology,2010,29(11):197-198.
[11]吴新根,葛家理.应用人工神经网络预测油田产量[J].石油勘探与开发,1994,21(3):75-78,131.WU Xingen,GE Jiali.The application of artificial neural network in predicting output of oil fields[J].Petroleum Exploration and Devlopment,1994,21(3):75-78,131.
[12]纪天亮,卢双舫,唐明明,等.致密油藏水平井压裂后产能预测方法[J].大庆石油地质与开发,2016,35(2):165-169.JI Tianliang,LU Shuangfang,TANG Mingming,et al.Productivity predicting method of the fractured horizontal wells in the tight oil reservoirs[J].Petroleum Geology&Oilfield Development in Daqing,2016,35(2):165-169.
[13]李克文,王义龙,苏玉亮.一种基于动态径向基函数神经网络的油田产量预测方法:CN104732303A[P].2015-06-24.LI Kewen,WANG Yilong,SU Yuliang.Oilfield production prediction method based on dynamic radial basis function neural network:CN104732303A[P].2015-06-24.
[14]RUMELHART D E,HINTON G E,WILLIAMS R J.Learning representations by back-propagating errors[J].Nature,1986,323:533-536.
[15]HOCHREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997,9(8):1 735-1 780.
[16]ECK D,SCHMIDHUBER J.Finding temporal structure in music:blues improvisation with LSTM recurrent networks[C].Martigny:Proceedings of the 12th IEEE workshop on neural networks for signal processing,2002.
[17]PéREZ-ORTIZ J A,GERS F A,Eck D,et al.Kalman filters improve LSTM network performance in problems unsolvable by traditional recurrent nets[J].Neural Networks,2003,16(2):241-250.
[18]范竣翔,李琦,朱亚杰,等.基于RNN的空气污染时空预报模型研究[J].测绘科学,2017,42(7):76-83.FAN Junxiang,LI Qi,ZHU Yajie,et al.A spatio-temporal prediction framework for air pollution based on deep RNN[J].Science of Surveying and Mapping,2017,42(7):76-83.