基于GWO-BiLSTM的岩性识别方法研究与应用  

Research and Application of Lithology Identification Method Based on GWO-BiLSTM

作　　者：崔文洁 赵军龙 陈家鑫[1,2] 张雨辰 孙婧 金利睿 CUI Wenjie;ZHAO Junlong;CHEN Jiaxin;ZHANG Yuchen;SUN Jing;JIN Liruin(Xi′a Shiyou University,Xi′an 710000,China)

机构地区：[1]西安石油大学地球科学与工程学院,陕西西安710000 [2]西安石油大学陕西省油气成藏地质学重点实验室,陕西西安710000

出　　处：《河北地质大学学报》2024年第5期30-37,共8页Journal of Hebei Geo University

基　　金：陕西省自然科学基础研究计划(2019JM-359)。

摘　　要：为解决常规岩性识别方法精度不高、耗时较长且受人为影响较大等问题,构建了基于GWO-BiLSTM的岩性识别方法。结合录井资料、岩心资料以及测井资料,采用常规方法对研究区进行岩性识别,效果较差,进而利用GWO-BiLSTM模型在研究区展开岩性识别工作。根据皮尔逊函数对各测井曲线与岩性进行分析,优选出相关系数绝对值大于0.3的测井曲线值作为输入特征,采用灰狼优化算法对BiLSTM超参数组合随机生成与更新,从而更加快速地获取最优解,进一步提高模型的效率以及准确率。实验表明,基于GWO-BiLSTM模型的岩性识别准确率达96%,与BiLSTM模型、RF模型、BP神经网络和SVM模型相比具有较高的准确率,验证了该模型在识别复杂岩性时的可靠性,并为复杂岩性识别提供了方法参考。In order to solve the problems of low accuracy,long time and human influence,a new lithology identification method based on GWO-BiLSTM is proposed.Combining well log data,core data and well logging data,the lithology identification in the study area is carried out by conventional method,but the effect is poor,and then the lithology identification is carried out by using the GWO-BiLSTM model.Each log and lithology were analyzed according to Pearson function,and log values with absolute correlation coefficients greater than 0.3 were selected as input features.Grey Wolf optimization algorithm was used to randomly generate and update BiLSTM hyperparameter combination,so as to obtain the optimal solution more quickly and further improve the efficiency and accuracy of the model.Experiments show that the accuracy rate of lithology identification based on GGO-BILSTM model is 96%,which is higher than that of BiLSTM model,RF model,BP neural network and SVM model,which verifies the reliability of this model in the identification of complex lithology,and provides a method reference for complex lithology identification.

关 键 词：复杂岩性识别 灰狼优化算法 双向长短时记忆神经网络 交会图法 机器学习 

分 类 号：P618.13[天文地球—矿床学]