BP方法及其在产油量中的应用模拟课题.doc

PAGE 摘 要 本课题是模拟课题。 准确预测油气井产量对指导油田开发实践具有重要意义。本文介绍了神经网络的基本内容，对其概念、框架、发展历史、计算和分类都进行了细致的描述；介绍了人工神经网络、模糊神经网络、自组神经网络和BP神经网络的基础知识；建立了利用BP神经网络预测油井产量的方法，并进行了实例计算。本文研究表明，神经网络通过模拟人脑的某些机理和机制而实现某个方面的功能，BP神经网络方法是一种有导师的学习算法，它具有自组织、自学习、自适应、容错和抗干扰能力等优越性；反向传播神经网络模型包括节点输出模型、作用函数模型、误差计算模型等；反向传播神经网络通过信息向前输入和误差反向传播，能得到较小的均方收敛误差，其缺点是收敛速度较慢；利用所建立的神经方法预测油井产量操作简便，精度较高。 关键词：油田开发；产油量；神经网络；数学模型 Abstract This topic is simulating thesis. Forecasting oil and gas well output accurately to instruct the oilfield development practice has vital significance. This paper introduced the basic content of including neural network, including its concept, frame, development history, computational method and its classification. At the same, this paper introduced elementary knowledge of neural network, fuzzy neural network, self-organization neural network and BP neural network, established the method of forecasting oilwell output using the BP neural network, and carried out an example run. The result indicates that, neural network can realize some kind of function through simulating certain mechanisms of human brain. BP neural network method is a kind of teachers learning algorithm. It has many priorities over other methods, such as self-organization, self-learning, self-adaptation, error-tolerance and good anti-disturbance ability. Back propagation neural network includes node-point output model, function model, error calculation model, etc.. Back propagation neural network can obtain relatively smaller mean square error through forward propagation of input and back propagation of error. Its shortcoming is the convergence rate is slow. Predicting the output of oilwell using the established neural approach is easy to operate and has high accuracy. Key words:oilfield development, production rate, neural network, mathematical model 目 录 TOC \o 1-3 \h \z \u HYPERLINK \l _Toc139329925 第1章 概 述 PAGEREF _Toc139329925 \h 1 HYPERLINK \l _Toc139329926 1.1 油田开发指标预测方法的发展历史及其多样性 PAGEREF _Toc139329926 \h 1 HYP