The natural gas generation process is simulated by heating source rocks of the Yacheng Formation, including the onshore-offshore mudstone and coal with kerogens of Type II_2-III in the Qiongdongnan Basin. The aim is t...The natural gas generation process is simulated by heating source rocks of the Yacheng Formation, including the onshore-offshore mudstone and coal with kerogens of Type II_2-III in the Qiongdongnan Basin. The aim is to quantify the natural gas generation from the Yacheng Formation and to evaluate the geological prediction and kinetic parameters using an optimization procedure based on the basin modeling of the shallow-water area. For this, the hydrocarbons produced have been grouped into four classes(C_1, C_2, C_3 and C_(4-6)). The results show that the onset temperature of methane generation is predicted to occur at 110℃ during the thermal history of sediments since 5.3 Ma by using data extrapolation. The hydrocarbon potential for ethane, propane and heavy gaseous hydrocarbons(C_(4-6)) is found to be almost exhausted at geological temperature of 200℃ when the transformation ratio(TR) is over 0.8, but for which methane is determined to be about 0.5 in the shallow-water area. In contrast, the end temperature of the methane generation in the deep-water area was over 300℃ with a TR over 0.8. It plays an important role in the natural gas exploration of the deep-water basin and other basins in the broad ocean areas of China. Therefore, the natural gas exploration for the deep-water area in the Qiongdongnan Basin shall first aim at the structural traps in the Ledong, Lingshui and Beijiao sags, and in the forward direction of the structure around the sags, and then gradually develop toward the non-structural trap in the deep-water area basin of the broad ocean areas of China.展开更多
现有的传统孔隙分割方法不能准确提取岩心图像中细小狭长的孔隙,且容易受到岩石图像噪声的干扰,针对上述问题,提出了一种深度学习网络模型ARC-Unet(Attention and Recurrent-Convolution Unet),用于更加精确的分割岩石孔隙。采用Unet作...现有的传统孔隙分割方法不能准确提取岩心图像中细小狭长的孔隙,且容易受到岩石图像噪声的干扰,针对上述问题,提出了一种深度学习网络模型ARC-Unet(Attention and Recurrent-Convolution Unet),用于更加精确的分割岩石孔隙。采用Unet作为基础网络,并在网络上加入注意力机制,用于解决在分割时小面积的孔隙容易被漏分割的情况。将循环卷积模块代替原来的卷积模块,可以拟合更多的岩石特征,提高孔隙分割的准确度。通过在采集并制作的岩石数据集上进行训练并在通过在测试集的分割结果上进行模型评估,改进模型在测试集上的F1达到了88.15%,有着较好的岩石孔隙分割结果。展开更多
基金The Western Light Talent Culture Project of the Chinese Academy of Sciences under contract No.Y404RC1the National Petroleum Major Projects of China under contract No.2016ZX05026-007-005+2 种基金the Key Laboratory of Petroleum Resources Research Fund of the Chinese Academy of Sciences under contract No.KFJJ2013-04the Science and Technology Program of Gansu Province under contract No.1501RJYA006the Key Laboratory Project of Gansu Province of China under contract No.1309RTSA041
文摘The natural gas generation process is simulated by heating source rocks of the Yacheng Formation, including the onshore-offshore mudstone and coal with kerogens of Type II_2-III in the Qiongdongnan Basin. The aim is to quantify the natural gas generation from the Yacheng Formation and to evaluate the geological prediction and kinetic parameters using an optimization procedure based on the basin modeling of the shallow-water area. For this, the hydrocarbons produced have been grouped into four classes(C_1, C_2, C_3 and C_(4-6)). The results show that the onset temperature of methane generation is predicted to occur at 110℃ during the thermal history of sediments since 5.3 Ma by using data extrapolation. The hydrocarbon potential for ethane, propane and heavy gaseous hydrocarbons(C_(4-6)) is found to be almost exhausted at geological temperature of 200℃ when the transformation ratio(TR) is over 0.8, but for which methane is determined to be about 0.5 in the shallow-water area. In contrast, the end temperature of the methane generation in the deep-water area was over 300℃ with a TR over 0.8. It plays an important role in the natural gas exploration of the deep-water basin and other basins in the broad ocean areas of China. Therefore, the natural gas exploration for the deep-water area in the Qiongdongnan Basin shall first aim at the structural traps in the Ledong, Lingshui and Beijiao sags, and in the forward direction of the structure around the sags, and then gradually develop toward the non-structural trap in the deep-water area basin of the broad ocean areas of China.
文摘现有的传统孔隙分割方法不能准确提取岩心图像中细小狭长的孔隙,且容易受到岩石图像噪声的干扰,针对上述问题,提出了一种深度学习网络模型ARC-Unet(Attention and Recurrent-Convolution Unet),用于更加精确的分割岩石孔隙。采用Unet作为基础网络,并在网络上加入注意力机制,用于解决在分割时小面积的孔隙容易被漏分割的情况。将循环卷积模块代替原来的卷积模块,可以拟合更多的岩石特征,提高孔隙分割的准确度。通过在采集并制作的岩石数据集上进行训练并在通过在测试集的分割结果上进行模型评估,改进模型在测试集上的F1达到了88.15%,有着较好的岩石孔隙分割结果。