基于时间序列BP神经网络的煤层气井排采制度优化  

Production systems optimization of a CBM well based on a time series BP neural network

作　　者：吴财芳[1,2] 姚帅[1,2] 杜严飞[1,2] 

机构地区：[1]中国矿业大学资源与地球科学学院,江苏徐州221008 [2]中国矿业大学煤层气资源与成藏过程教育部重点实验室,江苏徐州221008

出　　处：《中国矿业大学学报》2015年第1期64-69,共6页Journal of China University of Mining & Technology

基　　金：国家科技重大专项(2011ZX05034);国家自然科学基金项目(41272178);江苏省"青蓝工程";江苏高校优势学科建设工程项目(PAPD)

摘　　要：为了准确预测煤层气井产能,定量分析和优化煤层气井排采制度,利用神经网络具有的非线性映射能力和预测能力,基于时间序列思想构建了结构为14-13-7的BP神经网络煤层气井产能预测模型,在不同排采制度下对潘庄区块CM1井进行了未来7d的产能预测.结果表明:依据产水量、井底流压调控量度临界值分别为0.2 m3/d,0.1MPa,可以设计24种排采制度调整方案.其中,使产气量大幅减小的排采制度调整方案有5种,小幅减小的有7种,小幅增大的有7种,大幅增大的有5种.可以根据未来生产需要,采取不同的排采制度,对煤层气井产能实现人工实时调控.To accurately predict the productivity of a coalbed methane（CBM）well and quantitatively optimize the drainage systems of CBM wells,a CBM well productivity prediction model with a 14-13-7structure was constructed by the time series method with nonlinear mapping and predictive abilities of back-propagation（BP）neural networks.The future 7-day productivity of the CM1 well for different drainage systems in Pangzhuang Block were forecast by the model.The results indicate that 24 types of adjustment schemes for drainage systems are designed based on the water rate and the bottom hole flowing pressure regulation thresholds of 0.2m3/d and 0.1 MPa,respectively.Five types of schemes significantly reduced the gas yield,seven types of schemes slightly reduced the gas yield,seven types of schemes slightly increased the gas yield,and five types of schemes significantly increased the gas yield.Different drainage systems can be adopted to artificially regulate the productivity of CBM wells according to future yield requirements.

关 键 词：时间序列 BP神经网络 排采制度 煤层气井 优化 

分 类 号：P618.11[天文地球—矿床学]