【作者】 李鹏举；

【导师】 宋延杰；

【作者基本信息】 大庆石油学院 ， 地球探测与信息技术， 2003， 硕士

【摘要】 随着油气勘探和开发程度逐渐加深，迫切需要解决薄层划分和厚层细分问题，重新评价老井测井资料，寻找漏失掉的油气层。因此，有必要进行测井曲线的高分辨率处理，以提高薄层解释精度。 多极子阵列声波测井是目前各油田已推广使用的测井项目，现有的处理方法追求时差计算的精度，忽视了时差的纵向分辨率。另外，目前常规声波时差测井曲线的高分辨率处理方法虽然取得了一定的实际效果，但其建模困难，实际应用效果还不十分令人满意。 本文充分地调研了阵列声波测井分波提取和高分辨率处理的各种方法，在比较其优缺点的基础上，针对多极子阵列声波测井提出了相关—互功谱结合多炮点的处理方法。即采用相关—互功谱法处理多极子阵列声波测井的最短共发射和共接收子阵列(0.152米)提取其时差，然后把得到的跨同一深度地层的所有最短共发射和共接收子阵列的声波时差平均，从而在获得高分辨率声波时差曲线的同时保证时差的计算精度和可靠性。将该方法应用于大庆油田实际，处理了八口井的多极阵列声波测井资料，并将其成果与常规处理的结果及实测的高分辨率声波时差曲线进行对比分析，表明该方法能够可靠、有效地提高声波时差的纵向分辨率，具有较高的实用价值。 本文在分析了常规声波测井的纵向响应特征，深入研究了低分辨率声波时差曲线和高分辨率声波时差曲线之间的非常复杂的非线性映射关系的基础上，将关键井高、低分辨率的声波时差曲线作为学习样本，构建学习样本集，以人工神经网络为技术手段，建立反演预测模型，进而生成其它井的高分辨率声波时差曲线。选择了大庆油田某区块八口井作为建模和预测对象，进行了实际资料的处理。人工神经网络用于常规声波时差曲线的高分辨率处理从理论上是可行的，从实际应用上看，也取得了一定的效果。因此，作为常规测井曲线高分辨率处理的一种新的尝试，该方法具有广阔的应用前景。更多还原

【Abstract】 With the gradual development of exploration and exploitation for oil and gas, it is pressing to solve the problem of distinguishing thin bed, reevaluating old well logs and seeking neglected oil beds. Therefore, well log must be high resolution processed so as to enhance the accuracy of evaluating thin bed.Multipole array acoustic logging has been applied widely by oil fields at present. Existing array waveforms processing techniques pursue accuracy of each wave component’s slowness calculated, while neglecting its vertical resolution. In addition, since it is difficult to build model, the applied effect of existing high resolution processing methods for conventional sonic log is not completely satisfactory though it has acquired some effects.Many methods of wave component’s pick-up and high resolution processing for array acoustic log have been investigated fully in this paper. Based on comparison of these methods’ merits and demerits, the processing method of correlation cross power spectrum combined with multiple-shot is presented aiming at multipole array acoustic logging. It processes the shortest common-source subarrays and common-receiver subarrays(0.152m) to extract acoustic slowness with cross power spectrum method. Then all the slowness of the same depth interval formation spanned by all the shortest common-source subarrays and common-receiver subarrays is averaged to ensure the accuracy and reliability of the slowness computed while obtaining high vertical resolution slowness curve. Apply the technique to array acoustic logging data in Daqing Oil Field, process eight well’s data of array acoustic logging, and compare its results with the results through conventional method and the measured high resolution slowness log to analyze its applied effects. This shows that the method is able to enhance slowness’ vertical resolution reliably and effectively, so it has very high applied value.Based on analyzing the vertical response characteristic of conventional sonic logging, researching thoroughly the very complex nonlinear relationship between conventional sonic log with low resolution and slowness log with high resolution, forecast model is established using training samples consisting of key wells’ slowness curves with high, low vertical resolution to create the other well’s slowness curve with higher vertical resolution by means of Artificial Neural Networks(ANN) technique. Actual data have been processed selecting some eight wells in Daqing Oil Field as object of establishing model and forecasting. It is feasible to apply ANN technique to high resolution processing for conventional sonic log theoretically and has acquired some effects practically. Hence, it has wide applied perspective as a new attempt in high resolution processing for conventional logs.更多还原