神经网络专家系统在预测单井日产量上的应用研究

STUDY ON THE APPLICATION OF THE NEURAL NETWORK EXPERT SYSTEM TO THE PREDICTION OF DAILY OUTPUT PER WELL

彭敦陆 徐士进 王汝成 郭延军

PENG Dunlu XU Shijin WANG Rucheng GUO Yangjun

南京大学 地球科学系, 江苏 南京 210093

Department of Earth Sciences, Nanjing University, Nanjing, Jiangsu 210093, China

本文全面论述了神经网络专家系统及其开发步骤,并利用所开发出的神经网络专家系统,以胜利樊家油田10口已知油井的储层参数和原油参数及单井日产量为学习样本,成功地预测了2口预测井的单井日产量。说明利用神经网络专家系统可以克服常规的统计方法和模糊方法无法正确确定各参数隶属度及权重分配的缺点,并指出通过神经网络专家系统进行油田预测的准确、快速及有效性。

This paper gives an overall discussion on the Neural Network Expert System and its development steps.Taking the reservoir and crude oil parameters and daily output per well for 10 available oil wells in Shengli-Fanjia oil field for example,the Neural Network Expert System developed has been used successfully to predict the daily output per well of two forecasting wells.Therefore,the Neural Network Expert System can overcome the defects of conventional statistical method and fuzzy method who are unable to correctly determine the subordination degree and jurisdiction distribution of each parameter.It is also pointed out that the oil field prediction by the Neural Network Expert System is accurate,rapid and effective.

https://doi.org/10.11781/sysydz199902156