基于小波-神经网络的井眼轨迹预测数学模型研究  

Study on mathematical model of well bore locus prediction based on wavelet-neural network

作　　者：孟庆华[1] 刘清友[1] 

机构地区：[1]西南石油大学油气藏地质及开发工程国家重点实验室

出　　处：《机械设计》2008年第9期25-27,共3页Journal of Machine Design

基　　金：国家自然科学基金资助项目(50474040)、(50674078);高校博士点专项基金资助项目(20050615003)

摘　　要：以复杂钻井系统动力学为基础,从所建立的钻头位移模型仿真实验中提取数据,把所得的井眼轨迹时间序列进行小波分析重构,分别得到平稳的随机序列、周期序列和非平稳趋势序列;并把所得平稳随机序列、周期序列、非平稳趋势序列作为输入层,构造小波-神经网络井眼轨迹预测模型,充分发挥小波分析的多分辨率功能和人工神经网络的非线性逼近能力。通过实验验证,此模型能有效提高预测精度、延长预见时间,这将为复杂钻井系统井眼轨迹预测研究提供新的理论依据及方法。Take the complicated well drilling system as a toundation, picking up data from simulation test on the established model of aiguille displacement. Carrying on the wavelet analysis reconstruction on the gained well bore locus-time sequence, the smooth random sequence, periodical sequence and non-smooth trend sequence were gained respectively. And let the gained smooth random sequence, periodical sequence and non-smooth trend sequence be regarded as the input layer to construct the locus prediction model of wavelet-neural network so as to sufficiently bring the multi-resolution function of wavelet analysis and the ability of non-linear approach of artificial neural network into play. Through experimentation it is verified that this model could efficiently enhance the prediction accuracy and prolong the time of forecast, this would provide new theoretical basis and method for the prediction study on well bore locus of complex well drilling system.

关 键 词：井眼轨迹 预测数学模型 小波分析 神经网络 仿真 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]