摘要
为对煤层气井产能的实时动态监测和预测预报,基于时间序列预测思想构建了适合于煤层气井产能预测的BP神经网络模型.以潘庄CM1井为预测实例,分析表明:BP神经网络能够较为准确地预测出煤层气井未来30 d的产能变化,其产气量和产水量预测平均相对误差分别为1.35%和3.88%;与COMET3预测结果相比,BP神经网络短期产能预测精度高,能更好的反映出煤层气井产能变化趋势.
In order to real-time dynamic monitoring and forecasting the coalbed methane well productivity,so build the BP neural network model that based on time series prediction idea suitable for coalbed methane well productivity prediction.Use Panzhuang CM1 well for forecast instance,the results show that： this model can accurately predict the productivity change of the CBM wells in the next 30 days,the average relative error of gas production and water production forecast respectively 1.35% and 3.88%;Compared with COMET3 predictions,short-term production forecast of BP neural network is better,and better reflect the variation trends of coalbed methane well production.
出处
《辽宁工程技术大学学报（自然科学版）》
CAS
北大核心
2013年第4期493-498,共6页
Journal of Liaoning Technical University (Natural Science)
基金
国家科技重大专项课题资助项目(2011ZX05034)
国家"973"课题资助项目(2009CB219605)
国家自然科基金重点资助项目(40730422)
青年科学基金资助项目(40802032)
关键词
BP神经网络
煤层气井
COMET3
产能预测
产气量
产水量
时间序列
短期预测
BP neural network
coalbed methane well
COMET3
productivity prediction
gas production
water production
time series
short-term production
