基于XGBoost算法的地震多数据体识别含油河道砂体

准噶尔盆地东部侏罗系多发育陆相河道砂体,砂体厚度小且横向连续性差,具有“一砂一藏”的特征,油水关系复杂,油气预测难度大,单一的地球物理信息难以准确确定河道砂体的分布特征和孔隙流体性质。为此,提出采用XGBoost算法从叠后地震剖面、叠后纵波阻抗和叠前纵横波速度比3种地震数据体中提取含油河道砂体的特征,以有效描述这类河道砂体含油性的空间展布特征,降低了利用常规单一地震属性对含不同流体砂岩储集层预测的多解性。通过模型测试证明了该方法在识别陆相河道砂体的含油气性方面的可行性,并在实际应用中取得了较好的效果。

Using Multiple Seismic Data Volume to Identify Oil-Bearing Channel Sandbodies Based on XGBoost Algorithm

Continental channel sandbodies are usually developed in the Jurassic strata of the eastern Junggar basin, which are characterized by poor lateral continuity, thin thickness and one reservoir in one sandbody, leading to difficulties in hydrocarbon prediction. It is hard to accurately determine the channel sandbody distribution and pore fluid properties by using single geophysical information. Therefore, XGBoost algorithm is presented, by which the characteristics about the oil-bearing channel sandbody can be identified from several seismic data volume such as post-stack seismic section, post-stack P-wave impedance and pre-stack wave velocity ratio, and the distribution of the kind of oil-bearing channel sandbodies can be effectively described and multi-solution of the prediction of sandstone reservoirs containing multiple fluids by using conventional profiles of a single seismic attribute can be solved. The test of the model proves the feasibility of using the method to identify the oil-bearing properties of continental channel sandbodies and good effects have been gained in field application.