基于Kriging代理模型的造斜率预测方法研究

发布时间：2018-05-11 03:28

本文选题：Kriging代理模型 + 造斜率　；参考：《科学技术与工程》2017年03期

【摘要】：造斜率受导向工具结构、类型、钻井参数、井眼轨迹、地层和钻头等因素的影响,构成了一个多变量影响的非线性耦合体系。造斜率预测呈现一定的模糊性、随机性和非线性特点,难以用显性的定量关系式来精确刻画造斜率与其影响因素之间的函数关系。从回归分析预测法的角度,提出一种采用Kriging代理模型构建预测功能函数,进行造斜率预测的新方法。该方法基于空间插值理论,以造斜率预测值为输出目标。首先选取造斜率主要影响因素作为输入参数,确定训练样本和测试样本;然后,计算Kriging模型中的最优超参数,并构建该预测模型。最后,以测试样本为基准,计算模型预测性能指标,完成对Kriging代理模型的造斜率预测性能评价。通过现场案例研究表明,与常见的多元回归模型和径向基函数模型(radial basis function,RBF)相比,基于Kriging代理模型的造斜率预测方法在均方根误差(root-mean-square error,RMSE)、最大绝对误差(maximum absolute error,MAE)和平均绝对误差(average absolute error,AAE)3个常见的定量指标上表现了更佳的预测性能。预测结果更稳健,计算量更小,对造斜率的预测精度较高,能克服几何法预测精度不高或力学法计算量大等缺点。在工程应用中能节约计算成本和提高预测效率,便于推广应用。
[Abstract]:The slope rate is influenced by the structure, type, drilling parameters, borehole trajectory, formation and bit of the steering tool, and constitutes a multivariable nonlinear coupling system. The prediction of slope is fuzzy, stochastic and nonlinear. It is difficult to describe the functional relationship between slope making and its influencing factors by explicit quantitative relationship. From the point of view of regression analysis and prediction, a new method for slope prediction using Kriging agent model is proposed. This method is based on spatial interpolation theory and takes slope prediction as output target. Firstly, the main factors of slope construction are selected as input parameters to determine the training samples and test samples, and then the optimal superparameters in the Kriging model are calculated, and the prediction model is constructed. Finally, the slope prediction performance of Kriging proxy model is evaluated by calculating the performance index of the model based on the test samples. The field case study shows that, compared with the conventional multivariate regression model and radial basis function model, The slope prediction method based on Kriging agent model shows better prediction performance on three common quantitative indexes, root-mean-square error (RMSE), maximum absolute error (maximum absolute error mae) and average absolute error (average absolute error). The prediction results are more robust, the calculation amount is smaller, and the prediction accuracy of slope building rate is higher, which can overcome the shortcomings of low precision of geometric method and large amount of calculation of mechanical method. It can save calculation cost and improve prediction efficiency in engineering application, and is easy to be popularized and applied.
【作者单位】：长江大学机械工程学院;非常规油气湖北省协同创新中心;水电机械设备设计与维护湖北省重点实验室(三峡大学);
【基金】：石油天然气装备教育部重点实验室(西南石油大学)项目(OGE201403-01) 国家自然科学基金(51275057) 湖北省教育厅科学研究计划中青年人才项目(Q20151301) 水电机械设备设计与维护湖北省重点实验室开放基金(2016KJX12)资助
【分类号】：TE243

【相似文献】