石油与天然气地质 ›› 1999, Vol. 20 ›› Issue (2): 129-132.doi: 10.11743/ogg19990207

• 论文 • 上一篇    下一篇

神经网络专家系统在预测石油可采储量上的应用——以胜利樊家油区为例

彭敦陆, 徐士进, 王汝成, 郭延军   

  1. 南京大学地球科学系, 江苏南京 210093
  • 收稿日期:1998-12-27 出版日期:1999-06-25 发布日期:2012-01-18

APPLICATION OF NEURAL NETWORK EXPERT SYSTEM TO PREDICTING THE PRODUCIBLE OIL RESERVES—An example of Fanjia Oil district in Shengli Oilfields

Peng Dunlu, Xu Shijin, Wang Rucheng, Guo Yanjun   

  1. Department of Earth Sciences, Nanjing University, Nanjing
  • Received:1998-12-27 Online:1999-06-25 Published:2012-01-18

PDF (PC)

928

摘要：

影响石油可采储量的因素是多种多样的,很难用一简单的表达式来描述两者之间的关系。神经网络专家系统为解决这一问题提供了新途径。具体过程为(1)选取储层参数(累积厚度、温度、有效孔隙度、有效渗透率、压力)和原油参数(含油饱和度、地下原油粘度和密度)等8个参数作为特征参数;(2)将参数进行标准化和归一化处理;(3)以樊家油田8个已知采油区作为学习样本对网络进行训练;(4)运用已训练好的神经网络专家系统对未知油区进行预测。预测结果与实际情况吻合很好,误差都在允许范围之内。

关键词: 神经网络, 专家系统, 可采储量, B-P算法, 特征参数

Abstract:

There are various kinds of factors that effect producible oil reserves.It is difficult to describe the relationship between the producible oil reserves and the effecting factors by an expression simply.Neural network expert system provides a new approach to solve the problem.Particular process is like that:1. select five reservoir parameters (cumulative thickness,temperature,effective porosity and permeability,pressure) and three crude oil parameters (saturation,viscosity and density)as characteristic parameters;2.the parameters should be standardized and normalized;3.take 8 known oil recovery areas in Fanjia sub oil field as learning samples to train the network system;4.use the trained system to predict the producible oil reserves of unkown oil areas.The prediction results fitted well with actural circumstances and the errors are exceptable.

Key words: neural network, expert system, producible reserves, B-P algorism, characteristic parameter