基于GGA优化的PNN方法在气水储层识别中的应用

Application of PNN method based on GGA optimization in identification of gas-water zones

张光辉 周 文 张银德 金文辉 包 艳

成都理工大学信息管理学院，四川 成都 610059 成都理工大学能源学院，四川 成都 610059

College of Information and Management, Chengdu University of Technology, Chengdu 610059, China College of Energy Resources, Chengdu University of Technology, Chengdu 610059, China

能够较准确地进行气水层识别一直以来都是油气勘探开发非常关注的一个问题。在认真分析Z气田山2段储层地质特征的基础上，提出了基于格雷码遗传算法优化的概率神经网络方法（GGA-PNN），探索了该方法在气水层识别中的应用。首先综合常规测井资料和试气资料构建59个气水层样本（其中学习样本36个，预测样本23个），并进行数据归一化处理，然后采用格雷码遗传算法来训练PNN平滑参数和隐中心矢量建立起气水层目的层段识别模型。利用该模型对36个建模训练样本进行回判，正确率达100%。然后对23个预测样本进行识别，结果正确的有22个，预测精度达95.65%，其中一个误判样本是把干层判识为气层。由此表明，利用GGA-PNN方法对山2段未知流体属性的正确识别是可行的。

 The accurate identification of gas and water zone has been a problem which is focused on petroleum exploration and development. On the basis of analyzing the geologic characteristics of Ｓhan 2 reservoir in Z Gas Field, this paper proposes the PNN method based on GGA optimization with gray code and discusses its application in the identification of gas and water zones. Firstly, the conventional logging data was combined with gas testing data to set up the samples of 59 gas-water zones, including 36 studying samples and 23 predicting samples, and to normalize the data. Then the GGA was used to train the smooth parameter and hiden central vector to establish the gas-water zone identifying model for interest zone. The model was used to distinguish 36 samples with an accuracy degree of 100%. Then, 23 predicting samples were distinguished again, as a result, 22 predicting samples were correctly distinguished and the accuracy degree was 95.65%, in which a judgment error is that the dry zone was interpreted as gas zone. So it is feasible to correctly distinguish the uncertain fluid property of Ｓhan 2 reservoir by GGA-PNN method.