BP神经网络在致密砂岩储层识别中的应用

川西须家河组地层岩性复杂,属于超致密低孔渗储层,所以储层识别是该地层天然气勘探中所面临的关键问题和难点之一。针对常规储层识别准确率不高的状况,提出利用BP神经网络进行储层的气水干层识别,利用模糊聚类和产层测试结果标定建模样本,采取随机抽样形成建模集与测试集,建立BP神经网络模型对23口井的储层进行气水干层预测,正确率达77.9%以上,明显提高了该地区的测井解释精度,并提供了一种准确率较高的储层预测方法。

THE APPLICATION OF BP NEURAL NETWORK IN IDENTIFICATION OF THE COMPACT SANDSTONE RESERVOIR

The stratums of Xujiahe in west of Sichuan province,which belong to the compact and low porosity reservoir,are so complex that makes the reservoir identification of natural gas exploration in Sichuan basin be one of the key nodi. Considering the low identification accuracy of conventional well logging,the writer has adopted BP-neural network method (BP-NN),with stratum samples of well logging data,which is identified by fuzzy clustering method,to rec- ognize the gas,liquid and solid objects and obtain modeling samples and testing collection eventually.Then the BP- NN of stratum identification is established by the modeling samples and testing collection at random.The accuracy percent is more than 77.9% from 23 wells of Xujiahe,so the writer got the conclusion that the BP-NN has actually and greatly improved the interpretation precision.