基于GBDT算法的页岩储层脆性地震预测方法——以渝西足201井区龙马溪组页岩储层为例

Brittle Seismic Prediction Method of Shale Reservoir Based on GBDT Algorithm—Taking the Shale Reservoir of Longmaxi Formation in the 201 Well Area of Western Chongqing as an Example

摘要: 随着页岩气勘探开发的深入，页岩气储层的脆性评价显得越来越重要，脆性指数评估直接影响水平井钻探的效率和安全性，可靠的页岩脆性预测方法对提高页岩气开发效益具有着重要的意义。本研究以渝西足201井区龙马溪组页岩为例，首先基于Zoeppritz方程进行叠前地震反演获得纵、横波速度以及密度从而得到所需要的弹性参数地震反演体；利用岩石弹性参数实验数据、阵列声波测井数据及全岩X-射线衍射实验数据，经过校正及解释获得连续的矿物脆性指数数据及对应的岩石弹性参数数据作为机器学习算法的样本集；训练高效准确的学习方法和GBDT梯度决策树预测模型；采用叠前地震反演岩石弹性参数体及GBDT梯度决策树学习算法，实现研究区页岩储层脆性的智能化预测。

Abstract: With the deepening of shale gas exploration and development, the brittleness evaluation of shale gas reservoirs is becoming more and more important. The brittleness index evaluation directly affects the efficiency and safety of horizontal well drilling. Reliable shale brittleness prediction methods are of great significance to improve the efficiency of shale gas development. In this study, the shale of the Longmaxi Formation in the 201 well area of western Chongqing is taken as an example. Firstly, pre-stack seismic inversion is performed based on the Zoeppritz equation to obtain the P-wave, S-wave velocity and density to obtain the required elastic parameter seismic inversion body. Using the experimental data of rock elastic parameters, array acoustic logging data and whole rock X-ray diffraction experimental data, continuous mineral brittleness index data and corresponding rock elastic parameter data are obtained after correction and interpretation as the sample set of machine learning algorithm. Train efficient and accurate learning methods and GBDT gradient decision tree prediction models; the pre-stack seismic inversion of rock elastic parameters and GBDT gradient decision tree learning algorithm are used to realize the intelligent prediction of shale reservoir brittleness in the study area.