机器学习驱动的地震多属性分析表征扇三角洲沉积

识别扇三角洲沉积体对恢复古湖泊沉积体系和指导斜坡带岩性油气藏勘探有重要意义。以辽河东部凹陷铁匠炉地区沙一下亚段低渗透储层扇三角洲沉积为例,在岩心分析的基础上,通过单一地震属性分析、多属性聚类优选、机器学习等方法,构建了超限学习机算法驱动的地震多属性分析预测砂岩分布,进而识别扇三角洲沉积相展布的方法。研究表明:超限学习机算法在地震多属性预测砂岩分布方面具有良好的适用性;平均绝对振幅、能量半时和波谷数三种地震属性对砂岩分布表征有效;研究区沙一下亚段存在南物源,在古湖泊南岸发育一个小型扇三角洲,在古湖泊北岸发育两个大型扇三角洲,每个扇三角洲由两个朵叶体组成。研究为斜坡带扇三角洲砂体分布预测和沉积相表征提供了有效方法。

Sedimentary facies characterization of fan delta based on machine learning driven multi-seismic attributes analysis

The identification of fan delta is of great significance to the rebuilding of paleolake sedimentary system and to the exploration of lithologic reservoirs in slope zone. In this work, fan delta of lower part in 1st member of Shahejie Formation in Tiejianglu area, eastern sag of Liaohe Depression is taken as an example. On the basis of drilling core analysis, an extreme learning machine driven method of sandstone distribution prediction is built and then used in the characterization of fan delta sedimentary facies. The method is built based on a set of methods including single seismic attributes analysis, multi attribute clustering optimization, and machine learning. This study suggests that 1) extreme learning machine can be used in multi seismic attributes analysis to predict sandstone distribution; 2) three seismic attributes including average absolute amplitude, energy half time, and number of trough can be used as characteristic variables for sandstone distribution; 3) there is sediment-source from the south in the area which formed a small fan delta on the south bank of the paleolake; meanwhile, on the north bank of the paleolake in study area, there deposited two large fan deltas each of which are formed by two lobes. This study provides an effective method for the distribution prediction and sedimentary facies characterization of fan delta sandbodies in slope zone.