联合神经网络在储层参数预测中的研究与应用

Research on Committee Neural Network Model for Reservoir Physical Parameter Prediction

地质储层参数在建立地质模型中起着关键作用,储层参数通过井资料获得。常规测井解释中多通过经验公式或简化地质条件建立模型计算储层参数。提出了新的神经网络模型,基于BP神经网络、RBF神经网络、支持向量回归并通过单层感知器共同构成联合神经网络模型。该网络模型在储层参数预测过程中能针对单一神经网络的不足而自适应调节网络结构,使预测效果达到最优,避免了单一网络在参数预测时的缺点,提高了预测的准确性。选取了同一地区的3口油井进行训练和验证实验,实验结果表明,联合神经网络模型优于单一的人工神经网络模型。

Geological reservoir physical parameters are crucial for building the three-dimensional geological model, and reservoir physical parameters are often obtained from logging data. In conventional logging interpretation, reservoir physical parameters are calculated by empirical formula or simplified geological conditions. The development of new technology has brought a new way for the prediction of reservoir physical parameters. This paper presents committee neural network （CNN）, a new neural network model, which is based on BP neural network, RBF neural network, support vector regression and single layer perception. This model could adjust network structure automatically and get the optimal predicted value, which avoids the defects of individual neural network in parameters prediction and improves the accuracy of the prediction. The model is used and tested in three wells logging in the same area. One well is used to establish the CNN model, and two wells are used to assess the reliability of constructed CNN model. Results show that the CNN model is better than individual neural network model.