利用超限学习算法计算致密砂岩储层渗透率方法研究

The Computation of Tight Sandstone Reservoir Permeability by Using Extreme Learning Algorithm

摘要: 致密砂岩储层渗透率较小，且变化较大，难以准确计算。为了解决利用常规测井曲线进行致密砂岩储层渗透率计算的问题，提出了利用超限学习算法进行渗透率计算的方法。对研究区致密砂岩储层10块岩样进行建模，另37块岩样进行预测，并利用留一法进行最佳参数的确定。预测结果表明，超限学习算法能准确预测致密砂岩储层的渗透率，误差较低，且计算速度快，训练所需时间仅为0.0004 s。

Abstract: The permeability of tight sandstone reservoir was small and varied, so it was difficult to calculate accurately. In order to solve the problem of calculating the tight sandstone reservoirs permeability by using conventional logging data, a method to calculate the tight sandstone reservoirs permeability by using the extreme learning algorithm was proposed. Ten rock samples of tight sandstone reservoirs in the studied area were used for modeling. Another thirty-seven rock samples were used for prediction, and the best parameter was determined by the leave-one-out method. The prediction results show that the extreme learning algorithm is very effective in predicting the permeability of tight sandstone reservoirs. The prediction error is very low; the calculating speed is fast; and the training time is only 0.0004 s.