摘要
单井原油产量时间序列预测对把握单井的动态情况是非常有意义的,同时它也可以为油井工作状态评估和控制提供依据。针对单井原油产量时间序列的非周期性和非线性,需要利用黑箱模型对其进行建模与预测。本文研究了结合奇异谱分析和径向基函数神经网络对单井原油产量时间序列进行多步预测的方法,证明奇异谱分析方法可以有效地减少多步预测过程中的误差传递,从而提高多步预测的可靠性。利用实测数据对该方法进行了检验,证明该方法是有利于提高预测精度的。
It is meaningful for indentifying the dynamic for performance of wells if we can forecasting the time series of daily output per well. An time series forecasting model combing radial basis function neural networks and singular spectrum analysis is proposed. It is proved that the error transporting is reduced greatly with the application of SSA. According this model the evidence for evaluation and controlling the well can be achieved. Case experiments are conducted on this method, and the reliability of this method was tested.
出处
《微计算机信息》
2009年第16期306-308,共3页
Control & Automation
关键词
时闻序列
奇异谱分析
径向基函数神经网络
time series
singular spectrum analysis
radial basis function neural networks
