基于随机森林算法的AVO类型判别  

AVO type discrimination based on random forest algorithm

作　　者：刘浩男 文晓涛[1] 何健[1] 陈芊澍 张晓琦 LIU Haonan;WEN Xiaotao;HE Jian;CHEN Qianshu;ZHANG Xiaoqi(Chengdu University of Technology,Chengdu,Sichuang 610059,China)

机构地区：[1]成都理工大学,四川成都610059

出　　处：《中国海上油气》2020年第5期73-81,共9页China Offshore Oil and Gas

基　　金：国家自然科学基金项目“深层碳酸盐岩储层流体地震预测理论与方法(编号:U1562111)”;“基于频变信息的流体识别及流体可动性预测(编号:41774142)”部分研究成果。

摘　　要：AVO技术可用于含气储层的识别,对油气勘探具有重要意义。人工识别储层AVO类型人为干扰因素较大,识别精度较低且耗时较长。由此,本文引入随机森林算法,利用Bootstrap重复抽样及枝叶节点分裂等技术生成大量决策树分类器,通过统计所有决策树的分类结果实现对储层AVO类型的判别。首先,基于工区内测井数据建立速度密度模型;其次,利用Shuey近似公式计算AVO曲线并获得该曲线对应的拟合多项式;第三,根据拟合多项式提取形态特征参数作为随机森林算法的训练数据集输入参数,将人工AVO类型识别结果作为输出参数,训练并得到决策树分类器;最后,以实际叠前地震数据的AVO曲线特征参数为输入参数,通过随机森林决策树分类判别得到工区内储层AVO类型。通过与近似支持向量机算法的对比结果可以看出,两种算法对储层AVO类型判别结果相近,都具有较高的准确率,但相比之下随机森林算法所需特征属性较少,泛化性较强,具有更好的普适性。AVO technology can be used to identify gas-bearing reservoirs and is of great significance to oil and gas exploration.However,the manual identification of AVO type of reservoirs is prone to man-made interferences,has low identification accuracy and is time-consuming.Therefore,this paper introduces the random forest algorithm,uses Bootstrap repeated sampling and branch and leaf node splitting techniques to generate a large number of decision tree classifiers,and realizes the identification of the AVO type of the reservoir by summarizing the classification results of all decision trees.First,a velocity density model is established based on logging data in the work area.Second,Shuey approximation formula is used to calculate the AVO curve and obtain the fitting polynomial corresponding to the curve.Third,the morphological feature parameters are extracted according to the fitting polynomial as the input parameters of the training data set of the random forest algorithm,and the artificial AVO type recognition results are used as the output parameters to train and obtain a decision tree classifier.Finally,the characteristic parameters of the AVO curve of the actual pre-stack seismic data are used as input parameters,and the AVO type of the reservoir in the work area is obtained through the classification and discrimination of the random forest decision tree.By comparing the results with those of the approximate support vector machine algorithm,it can be seen that the two algorithms are similar in terms of discrimination results of AVO type of the reservoir,and both have high accuracy,but the random forest algorithm requires fewer feature attributes,shows stronger generalization and has better universality.

关 键 词：AVO类型 随机森林 储层预测 分类判别 形态特征参数 

分 类 号：TE132.14[石油与天然气工程—油气勘探]