摘要
储层物性参数的估算是油气勘探开发的一个重要环节,然而物性参数的横向变化很难从分布稀疏的井孔资料插值外推得到精确描述,综合利用地震和测井资料可得到较精确的物性分布图像。传统地,将地震和测井资料综合来进行储层参数描述的方法有线性回归法,这种方法由于假定函数关系简单,因而很难得到准确的物性参数。后来Doyen〔5〕发展了协同克里金法,这种方法可得到较准确的物性参数,但处理过程复杂,需作大量的统计研究。人工神经网络是一种高度非线性系统,可逼进任意函数,因此可以综合地震和测井资料,利用人工神经网络来描述储层参数。我们采用前馈神经网络模型,包含输入层、输出层和隐伏中间层,隐伏中间层的个数以及各隐伏中间层的单元数可调,输出层为测井导出的和实测的储层物性参数集,输入层为地震数据集另加位置坐标。为保证全局最优,我们采用模拟退火算法来训练网络。用文献〔5〕的例子进行了试验,表明该方法其结果与协同克里金法相当,但处理过程却简单得多。
Estimating petrophysical parameter is an important link in oil & gas exploration and development,the lateral change of the petrophysical parameter,however,is very difficult to be accurately described by use of the interpolation and extrapolation of the sparsely distributed well data. An accurate petrophysical distribution image can be obtained by comprehensively applying the seismic and logging data.A traditional method of reservoir parameter description by synthesizing the seismic and logging data is linear regression method,by which it is very difficult to get the accurate petrophysical parameters,because of supposing its functional ralation to be simple.The synergetic Kriging technique was developed by Doyen afterwards,by which the relatively accurate petrophysical parameters may be obtained.But its handling procedure is complex and a great deal of statistical research has to be done.Artificial nerve network is a highly nonlinear system being able to approach any function,therefore the reservoir parameters can be described by it through synthesizing the seismic and logging data.The feedforward nerve network model,including input layer,output layer and concealed intermediate layer was adopted.The quantities of the concealed intermediate layer and its unit can be adjusted;the output layer is the petrophysical parameter set measured and derived from logging data;and the input layer is the seismic data set and position coordinates.In order to assure overall optimization,the simulation annealing algorithm was adopted to train the network.An experiment was carried out by use of the example in the fifth reference.The result indicates that such a method is in correspondence with the synergetic Kriging technique,but its handling procedure is much simple.
出处
《天然气工业》
EI
CAS
CSCD
北大核心
1997年第6期26-28,共3页
Natural Gas Industry
基金
国家自然科学基金
关键词
神经网络
地层参数
油气勘探
储层参数
Seismic data,Log data,Nerve network,Formation parameter,Reservoir description.