旋转磨料射流套管开窗预测模型及应用

Prediction model of casing window with swirling abrasive jet and its application

 刘康乐¹ 魏艳¹ 李莎¹ 王方祥²

天津石油职业技术学院石油工程系，天津 301607 中国石油大学（华东）石油工程学院，山东 青岛 266580）

Department of Petroleum Engineering, Tianjin Petroleum Vocational and Technical College, Tianjin 301607, China College of Petroleum Engineering, China University of Petroleum, Qingdao 266580, China)

 套管开窗径向水平井技术有助于提高低渗透储层的油气采收率，旋转磨料射流套管开窗方式是该技术新的发展方向，套管开窗直径和开窗深度是评价开窗效果的2个重要指标。由于无法实时监测井下开窗过程，需对开窗效果的预测进行研究。但套管开窗效果与旋转磨料射流参数之间存在复杂的非线性关系，使得传统预测方法的精度偏低。为此，文中提出了一种基于基因表达式编程建立套管开窗预测模型的新方法。通过遗传编码、遗传操作和适应度评估，建立了开窗直径和深度的直观数学表达式预测模型，利用套管开窗综合实验结果对该方法的准确性、实用性进行了验证。结果表明：该方法的预测精度较高，与检验样本相比，开窗直径和深度的平均误差分别为5.5%和4.4%；将预测模型的计算结果与套管开窗综合实验结果相比，开窗时间的误差为6.1%，开窗直径的误差为4.5%。因此，基因表达式编程算法能够更精确地推导出反映实际数据的最佳拟合函数，建立的预测模型可用于指导旋转磨料射流套管开窗现场施工。

 Casing window technology can help improve oil and gas recovery for low permeability reservoir, and casing window with swirling abrasive jet is a new development direction of it. The diameter and depth of the casing windows are two key parameters to evaluate the results. Due to disability of real-time monitoring of down hole casing window, it demands to study the prediction method of the effect of casing window. As a result of the complex non-linear relationship between the effect of casing window and the parameters of swirling abrasive jet, the accuracy of traditional prediction method is pretty low. Therefore, a new technique based on a gene expression programming(GEP） is presented. The genetic code, genetic operation and fitness evaluation were used to establish the intuitive mathematical expression prediction model of the diameter and depth of the casing windows and verified the accuracy and practicality of the method. Results show that the method has high accuracy, and compared with the sample, the average deviation of the diameter and depth of the casing windows are 5.5% and 4.4%. Comparing the result of prediction model and casing window comprehensive experiment, the deviations of casing window time and casing window diameter are 6.1% and 4.5% respectively. Consequently, gene expression programming can accurately obtain the best fitting function and the prediction model can be used for the guidance of site operation of swirling abrasive jet casing window.

10.6056/dkyqt201704031