摘要
新疆塔河地区地震属性参数与测井数据并非存在着明显的一一对应关系,很难用精确的算法来描述,使用多参数共同解释又往往产生相互矛盾的结果。针对这种情况,介绍一种三维地震资料多参数神经网络技术将丰富的地震资料与较稀少的钻井、测井数据结合起来进行沿层含油气性预测与评价的方法。首先分析塔河地区的地质规律及目的层段井资料的含油气性,研究井位处地震属性参数与目的层含油气性之间的关系,然后筛选出较稳定的、灵敏度较高的瞬时频率、反射强度、瞬时相位等优势地震参数进行归一化作为网络的输入数据,目的层段的测井或岩心油气显示为神经网络的期望值进行网络训练,最后用经过训练的网络对整个工区内各道相应目的层段进行沿层预测,形成油气预测图,实际应用表明该方法是可行的。
There is no explicit correlation between seismic attributes and logging data in Tahe oilfield, and interpretation using different attributes simultaneously often yields contradictory results. To circumvent the above difficulty, this paper proposes a multiple seismic attribute neural network approach. The approach combines relatively redundant seismic data with sparse drilling and logging data to predict and evaluate hydrocarbon potential along horizon. Firstly, analyze the geology in Tahe area and the hydrocarbon potential of the target by well data, establish the correlation between seismic attributes and the hydrocarbon potential of the target in the vicinity of wells, and single out some steady and sensitive dominant seismic attributes such as instantaneous frequency, reflection amplitude, and instantaneous phase as the input to the neural network after these attributes have been normalized. The hydrocarbon indications from logging or core data of the target were used to supervise the training of the neural network Finally, prediction of hydrocarbon potential was performed along the target for all traces in the survey area and areal hydrocarbon distribution was generated. The method was demonstrated on real data.
出处
《勘探地球物理进展》
2005年第4期290-293,共4页
Progress in Exploration Geophysics
关键词
沿层含油气性预测与评价
塔里木
塔河油田
神经网络
prediction and evaluation of hydrocarbon potential along horizon
Tarim
Tahe oilfield
neural network
