摘要
油田产量的预测一直是石油工作者研究的重要课题.针对油田产油量、产水量、地层压力和时间之间有着混沌的特征,利用多变量混沌时间序列等方法研究了油田产量的混沌建模和预测问题.用C-C算法确定每一个变量的嵌入维数和延迟时间,重构多元混沌时间序列的相空间;使用基于奇异值分解的主成分分析消除重构相空间的冗余变量和噪声干扰,建立了有较好泛化性能的多元混沌时间序列油田产量预测模型;最后将混沌时间序列预测和Elman神经网络进行耦合,创建了基于主成分分析前馈网络的多元混沌时间序列油田产量预测方法.应研究表明,提出的多变量混沌时间序列预测方法的预测精确度优于单变量预测,它可用于解决具有多变量混沌时间序列的预测问题.
Prediction of oil output is always an important hot topic of oil workers researching. Aiming at the characteristics of chaos among oil production, water production, formation pressure and time, the problems of chaos modeling and predicting about oil output were studied by multivariate chaotic time series method etc. In this thesis, we firstly used C-C algorithm to determine the embedding dimension and delay time of each variable to reconstruct the phase space of multivariate chaotic time series; then applied principal component analysis based on singular value decomposition to eliminate redundant variables and noise interference of reconstructing phase space, established the multivariate chaotic time series prediction model of oilfield production with good generalization performance; finally coupled the prediction method of chaotic time series with Elman neural network to create multivariate chaotic time series prediction method of oil field production based on PCA feed forward network. Application study shows that the forecasting accuracy of the novel method proposed is superior to that of single variable, which can be used to resolve prediction problem of multivariate chaotic time series.
出处
《数学的实践与认识》
北大核心
2016年第6期99-105,共7页
Mathematics in Practice and Theory
基金
四川省教育厅重点自然科学基金(11ZA024)
关键词
多变量混沌时间序列
C-C算法
相空间重构
主成分分析
ELMAN神经网络
multivariate chaotic time series
C-C algorithm
phase space reconstruction principal component analysis
Elman neural network