基于岩性识别的BP神经网络孔隙度预测

研究区目的层有效储层岩性包括细砂岩和粉砂岩两种,目前依赖于常规线性拟合方程在该区的孔隙度等参数预测中存在着较大的误差。为实现对储层参数的准确预测,结合了多种测井信息,在利用Fisher判别法对岩性进行识别的基础上,分别建立了细砂岩和粉砂岩的神经网络孔隙度预测模型。实际应用表明,分岩性所建立的非线性人工神经网络模型比常规线性模型具有更高的预测精度,能为该区后续的储层综合评价提供可靠的数据基础。

Prediction of porosity by BP neural network based on lithology recognition

The effective reservoir lithology of the study area includes fine sandstone and siltstone, and there is a large error in the prediction of porosity and other parameters in this area relying on conventional linear fitting equations. In order to achieve accurate prediction of reservoir parameters, combined with a variety of well logging information, the neural network porosity prediction models for fine sandstone and siltstone were established on the basis of Fisher’s discriminant method for identifying lithology. The practical application shows that the nonlinear artificial neural network model established by the lithology has a higher prediction accuracy than the conventional linear model, and can provide a reliable data foundation for the subsequent comprehensive evaluation of the reservoir in the area.