基于数据挖掘技术的深井钻速预测方法研究

摘要

为了对钻井过程中的机械钻速进行预测,提出了层次分析-神经网络组合模型,将层次分析法确定的钻速影响因素权重和原始数据一并作为输入信息,经过神经网络训练、迭代,确定已钻井钻速模型,最后利用模型对同一区域待钻井钻速进行仿真预测。仿真结果表明,应用层次分析法与前馈神经网络原理建立的AHP-BP组合模型在一定程度上可以提高迭代收敛速度和训练结果的可靠性;利用数据挖掘技术对钻速进行仿真预测,并与实钻值进行对比,可以对钻井状态进行评价,为优选钻进参数、钻头型号及钻井优化提供辅助决策支持。%To predict the rate of penetration（ROP） in the process of drilling,the combined model of analytic hierarchy process（AHP） and neural network was formulated.The weight of the ROP influencing factors determined by the AHP and the original data were taken as the input information.The ROP model of drilled well was determined through network training and iteration.Finally,the model was adopted to make a simulation prediction of the ROP of impending drilling well in the same area.The simulation findings show that the AHP-BP model established by the AHP method and feedforward neural network principle can,to some extent,improve the iterative convergence rate as well as the reliability of the training result.The data mining technology is adopted to make a simulation prediction of ROP which is then compared with the actual drilling value.In this way the drilling state can be assessed and the auxiliary decision support is provided for optimizing drilling parameters and bit types as well as drilling optimization.