基于地质信息约束的概率神经网络地震反射模式预测方法  

A method under the constraint of geological information for prediction of seismic reflection patterns with probabilistic neural network

作　　者：丁峰[1] 胡光义 尹成[1] 范廷恩 罗浩然[1] 张栋 

机构地区：[1]西南石油大学地球科学与技术学院,四川成都610500 [2]海洋石油高效开发国家重点实验室,北京100028 [3]中海油研究总院有限责任公司,北京100028

出　　处：《中国海上油气》2018年第1期127-131,共5页China Offshore Oil and Gas

基　　金：“十二五”国家科技重大专项“海上开发地震关键技术及应用研究(编号：2011ZX05024-001)”部分研究成果.

摘　　要：已有神经网络地震反射模式预测所得结果往往不稳定,多次预测可能出现多种结果,难以评价预测结果的准确性;而且预测值只代表了不同的反射类别,并没有相应的地质指示和含义。通过构建地质信息库和地质模式-地震模式的转换技术,建立了一种地质信息约束下的概率神经网络地震反射模式预测方法,解决了两个关键性问题,一是地质信息如何体现对地震信息的约束,二是地质信息约束的地震反射模式的高精度预测如何实现。以渤海秦皇岛32-6油田某河流相储层为例,选取20口井作为检验井,25口井作为约束井,本文方法储层地震反射模式预测的符合率为86%;模型及实际数据应用表明,在地震资料主频为45 Hz左右时,本文方法可以有效识别出3~8 m的河道砂体,并能够辨别大于2 m的泥岩隔夹层个数,具有高效、稳定、智能化、可解读性强等特点。At present,the method to predict seismic reflection patterns with neural network generally fails to keep outputs constant.The outputs vary and hence the accuracy of prediction cannot be guaranteed.Besides,the prediction values can only indicate reflection types and fail to reveal corresponding geological indications and meanings.Therefore,a method under the constraint of geological information for the prediction of seismic reflection patterns with probabilistic neural network has been built up based on the constructed geological information database and geological-to-seismic transformation technology,which can solve two critical problems about,firstly,how to reflect the constraint of geological information on seismic information,and secondly,how to accomplish the high-accuracy prediction of seismic reflection patterns under the constraint of geological information.In this paper,with a certain fluvial reservoir in the QHD32-6 oilfield of the Bohai sea as an example,where 20 wells were chosen for tests and another 25 wells for constraint,the coincidence rate for the prediction of reservoir seismic reflection patterns under the above method reached up to 86%.As indicated by simulation and in-field application,when the main seismic frequency is about 45 Hz,with the above method,the channel sand bodies of 3~8 m in thickness can be effectively identified and the number of mudstone interlayers thicker than 2 m distinguished.Therefore,the method is thought characterized by high effectiveness,stabilization,intellectualization,great interpretability,etc..

关 键 词：地质信息约束 质模式 震反射模式 震属性 率神经网络 皇岛32-6油田 

分 类 号：TE121[石油与天然气工程—油气勘探]