【作者】 刘玉霞；

【导师】 王延江；

【作者基本信息】 中国石油大学 ， 信号与信息处理， 2007， 硕士

【摘要】 在水平井钻井中,如何确定目标层的位置保证钻头在目标层内钻进是开发复杂储藏及薄油藏的一个非常关键的问题,也是目前我国地质导向钻井中一个急需解决的技术难点。本文以中石化重点攻关项目“地质导向钻井工艺技术研究”为背景,重点对地质导向钻井中随钻预测方法进行研究,分别利用支撑向量机、卡尔曼滤波理论、时间序列等理论分析综合测井参数、钻井参数及MWD随钻测量信息对钻头处的地质参数及井眼轨迹进行预测,以便进行实时地质导向。本文主要贡献如下:1、提出了基于支撑向量机统计学习理论的自然伽马随钻预测及校正方法,并进一步对地层电阻率进行了预测。试验表明,该方法能够及时有效的预测出钻头处的自然伽马和电阻率曲线,为井眼在产层中延伸提供了依据。2、根据参数估计理论,建立了基于卡尔曼滤波器的井眼轨迹预测模型,该模型可用于预测钻头处的井斜角和方位角,实验结果说明了该模型的有效性。3、将时间序列理论和支撑向量机相结合,建立了基于支撑向量机的时间序列预测模型,它通过机器学习的方法根据已钻井的轨迹参数构造支撑向量机进行井眼轨迹预测。实例分析表明,该方法简单有效且估计精度远远高于传统的定曲率法。更多还原

【Abstract】 During the horizontal well drilling, it is very important to know theposition of the geological target and maintain the well path at a constantrelative position within the target while developing complicated reservoirs orthin reservoirs, which is also a key technique problem of geosteeringdemanding an immediate solution at present in china.Based on the key project of SinoPec---’Research on GeosteeringTechniques’, this thesis mainly studies on the methods of predicting whilegeosteering drilling. After analysis of the well-logging parameter, drillingparameter and MWD data, some new methods are proposed to predict thegeological data and the trajectory data at the bit by some new informationprocessing techniques, and they can be used to geosteer the well whiledrilling in real time.The main contributions are as follows:1. A new method for predicting the gamma ray at the bit position isproposed based on support vector machine (SVM).The predicted gamma raycurve is then corrected by the data measured by LWD sensor along the wellpath, and the resistivity curve is predicted also. Experimental results show theproposed method can estimate the gamma ray and the resistivity curve at the bit effectively, and it can be used to geosteer the well while drilling in realtime.2. A novel wellbore trajectory prediction model is founded using thekalman filter, which is based on parameter estimation theory. It is used topredict the well inclination and azimuth angles. The experimental resultsshow that the proposed model is effective.3. Another well trajectory prediction model is established based on thecombination of time series and the SVM. The basic idea of this method is topredict the wellbore trajectory by learning using the trajectory data measuredby MWD. The experimental results show that the proposed method hashigher precision compared with the traditional methods and can be easilyused.更多还原