基于遗传算法的BP神经网络预测石油单井产量  

BP Neural Network Based on Genetic Algorithm to Predict Oil Single Well Production

作　　者：杜航 张涛[1] 陈岩[1] 赵银明[1] DU Hang;ZHANG Tao;CHEN Yan;ZHAO Yinming(School of Information and Mathematics,Yangtze University,Jingzhou 434023,China)

机构地区：[1]长江大学信息与数学学院,湖北荆州434023

出　　处：《成都大学学报（自然科学版）》2021年第1期57-61,70,共6页Journal of Chengdu University（Natural Science Edition）

基　　金：国家自然科学基金项目(61673006)。

摘　　要：针对传统预测方法建模复杂且预测精度低、传统BP神经网络算法收敛慢和可能陷入局部最优的问题,提出一种基于遗传算法的反向传播神经网络(BPANN-GA)用以预测单井产量,并进行实例分析,获得了较高精度的预测结果.该算法简单实用,预测精度较高,收敛快且避免了陷入局部最优,对石油单井产量预测工作有一定的意义.The prediction of oil single well production is of great significance to the sustainable and stable development of oil field production.In this paper,a back propagation neural network(BPANN-GA)based on genetic algorithm is proposed to predict single well production,aiming to deal with the problems in traditional prediction methods such as complex modeling and low prediction accuracy.Meanwhile,the traditional BP neural network algorithm is characterized by its slow convergence and the possibility of falling into local optimum.In addition,instance analysis is done on the newly proposed method and comparatively higher accuracy for prediction can be obtained.It can be concluded that the algorithm is simple and practical,with high prediction accuracy,fast convergence,and no possibility of falling into local optimum,which has certain significance for the prediction of single well production.

关 键 词：单井产量预测 BP神经网络 遗传算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TE331.1[自动化与计算机技术—控制科学与工程]