基于机器学习的孔隙度预测方法研究

Research on the porosity prediction method based on machine learning

介绍并比较了两种有监督的机器学习方法:BP神经网络和决策树。用两种方法分别论证了如何利用测井信息非线性地表示孔隙度。与传统的线性回归方法相比,机器学习效果更好,准确性更高。BP神经网络和决策树的应用效果表明,机器学习可以有效预测孔隙度,也可以应用于储层孔隙度预测中。相比之下,神经网络具有更高的准确性和更广阔的前景。

Two supervised machine learning methods, BP neural network and decision tree, are introduced and compared. Two methods were used to demonstrate how to use the logging information to represent porosity nonlinearly. Compared with the traditional linear regression method, the result from the machine learning is better and the accuracy is higher. The special effects of BP neural network and decision tree indicate that the porosity could be predicted effectively by the machine learning and can be applied to reservoir porosity prediction. In contrast, the neural networks have higher accuracy and broader prospects.