储层潜在敏感性B-P网络快速预测方法研究

对神经网络算法以及利用神经网络进行敏感性预测中经常出现的问题进行了探讨。着重介绍了合理网络结构的确定方法、学习样本的搜集与处理以及学习算法的改进等 ,并建立了改进的B P算法。通过水敏性预测 ,介绍了神经网络在敏感性预测中的具体应用过程 ,并对其在实际生产中的应用效果进行了验证。结果表明 ,利用改进后的神经网络对储层敏感性进行预测的方法 ,具有预测精度高、自适应好、容错能力强等优点。为了便于使用 ,编制了功能完备的敏感性预测软件 ,现场使用效果良好。

PREDICTION METHOD OF POTENTIAL SENSIBILITY OF FORMATION DAMAGE WITH B-P network

The artificial neural network is a new intelligent method developed in recent years. Some problems emergenced in predicting the potential damage of formation by the arithmetic of the artificial neural network are researched.Such problems include develompent of an reasonable network architecture, collection and handling of the samples, advance of the arithmetic. Taking the water sensitivity prediction as an example,the effect of prediction in the production is validated. The application result of this method shows that it has such advantages of high exactness, self adaptability and self organization. For the user's convenience,the software of prediction has been programmed. The field application of the software shows good effects.