新场气田沙溪庙组气藏单井产能预测  

SINGLE WELL PRODUCTIVITY PREDICTION OF SHAXIMIAO FORMATION RESERVOIR IN XINCHANG GAS FIELD

作　　者：蔡左花[1] 匡建超[2] 曾剑毅[1] 庞河清[1] 黄建红[1] 

机构地区：[1]成都理工大学能源学院 [2]成都理工大学商学院

出　　处：《钻采工艺》2009年第3期34-37,125-126,共4页Drilling & Production Technology

摘　　要：川西新场气田沙溪庙组气藏是典型的致密低渗透碎屑岩气藏,气藏储量丰富,可单井产能却很低。产能预测是编制气田开发规划部署、进行开发方案设计、开发动态分析、气井配产及开发方案调整的重要内容,但迄今为止,致密低渗储层的产能预测却仍是当下公认难点问题。针对沙溪庙组特定的地质特征,本文采用了最佳子集及GA-BP神经网络模型预测产能。通过最佳子集模型运算,获得与储层产能相关性最好的6个特征参数(分别是Φ、Ac、Φf、ΔR、F1、F2),再用GA-BP神经网络构建的储层产能预测模型,预测结果是绝对误差最大为0.98,最小为0.008,平均为0.036,相对误差最大为5.36%,最小为0.805%,平均为2.85%,说明所构建的基于最佳子集及GA-BP神经网络的储层产能预测模型预测结果理想,可以用于同类储层的产能预测。The gas reservoir of Shaximiao formation in Xinchang gas field of west Sichuan is a typical low-permeability tight clastic gas reservoir,the gas reservoir reserve is rich,but the deliverability is low.Fluids rate forecast is an important part of the gas field development planning to deploy,develop programs design,develop dynamic analysis of the allocation of gas production and development programs adjustment,but so far,low-permeability reservoir fluids rate forecast is still recognized as current difficult problems.Based on the specific geological features of Shaximiao formation,this article used the best subset and GA-BP neural network models to predict the deliverability.Through computing the best subset model,and then obtained the best features of 6 parameters(namely,(、Ac、(f 、(R、F1、F2)which have the best correlation with reservoir deliverability,and then used GA-BP neural network to build the reservoir fluids rate forecast models.The results of the forecast is the largest absolute error 0.98,0.008 minimum,with an average of 0.036,the relative error up to 5.36 %,0.805 % for the smallest,with an average of 2.85%,which present the prediction results of the reservoir productivity prediction model that Based on the best subset and GA-BP neural network are satisfactory,the forecast model can be used in a similar reservoir fluids rate forecast.

关 键 词：新场气田 沙溪庙组气藏 产能预测 最佳子集 神经网络 ’ 

分 类 号：TE328[石油与天然气工程—油气田开发工程]