基于支持向量机的钻井风险实时预测方法

DRILLING RISK REAL-TIME PREDICTIVE BASED ON SUPPORT VECTOR MACHINES

钻井过程具有高度的复杂性、井下状态的不确定性以及一些参数测量的不准确性,导致对于钻井过程的状态和潜在的风险无法及时作出正确的判断,有时甚至导致重大的事故。文中提出一种基于支持向量机的钻井过程风险预测方法。统计学习理论是基于小样本的统计学习方法。支持向量机是统计学习理论中最有效的方法,可以通过较小的样本建立准确度较高的数学模型,有很强的现实意义。通过支持向量机方法将钻井过程中实测的各种参数进行数据融合并训练相应的风险预测数学模型,通过此模型实现钻井过程中实时状态监测的目的。以某钻井过程的16个数据样本为例,其中8个数据样本作为训练样本,8个数据样本作为测试样本,验证结果说明了文章提出的基于支持向量机的钻井过程风险预测方法的有效性。

For there are higher complexity,uncertainty of underground states and inaccuracy of some measured parameter in the drilling process,the potential risk can not be judged accurately in time,and it even will cause drilling accident.In this paper,a drilling risk predictive method based on support vector machines was presented.Statistical learning theory was a small-sample statistical method,support vector machines was the most effective method in the statistical learning theory which could get accurate mathematical model though small sample,it has very strong practical significance.The support vector machine was used in data fusion,and the drilling risk predication model was trained by the measured data in drilling process.This model could be used to realize state monitoring in the process of drilling.Sixteen samples in the drilling process were used here,eight of it as training sample,and the other as test sample.Simulation results showed the proposed drilling risk predictive method based on support vector machines was effective.