基于双向长短期记忆网络的流体高精度识别新方法

A new method for high-precision fluid identification in bidirectional long short-term memory network

碳酸盐岩储层的储集空间类型多样、储层性质复杂,导致流体的测井响应受到强非均质性的影响,给流体识别工作带来极大困难。针对该问题,提出基于测井序列信息的双向长短期记忆网络(Bi-LSTM)流体识别模型,从测井响应特征差异性分析及相似性分析两方面出发,确定敏感曲线,结合Bi-LSTM网络的输入要求,建立流体识别样本库,并获得基于Bi-LSTM的流体识别模型。应用该方法对鄂尔多斯盆地马家沟组进行流体识别,与单向LSTM模型及其他3类机器学习算法预测结果进行对比。结果表明基于Bi-LSTM的流体识别模型流体识别的符合率从82.7%提高到91.5%,取得较好的应用效果;该模型既能充分利用井下对应深度测井曲线的响应值,又能兼顾测井曲线随深度的变化趋势和前后关联,最大程度避免储层纵向非均质性带来的影响,提高流体识别能力。

Due to the diversity of space types and complex reservoir properties of carbonate reservoir,the logging response of fluid is affected by strong heterogeneity,which causes great difficulties in fluid identification.Aiming at this problem of fluid identification in complex carbonate reservoir,the bidirectional long short-term memory network(Bi-LSTM)fluid identification model based on logging sequence information was proposed.From the analysis of the differences of logging response characteristics and the similarity analysis of characteristic curves,the sensitivity curve was determined.Combined with the input requirements of the Bi-LSTM,the fluid identification sample database was established,and the fluid identification model was obtained.This method was applied to identify Majiagou Formation in Ordos Basin.Compared with the predictions of the LSTM model and other three types of machine learning algorithms,the accuracy of fluid identification increased from 82.7%to 91.5%.The model can not only make full use of the logging response value of the corresponding depth,but also take into account the changing trend and correlation of the logging curve with depth in ways that avoid the influence of vertical heterogeneity of reservoir and therefore improve fluid identification ability.