基于机器学习的储层测井评价研究进展

Research progress of reservoir logging evaluation based on machine learning

储层测井评价的核心是数据分析和模型驱动方法的数学建模问题,将测井评价过程转化为机器学习过程,是提高储层测井评价自动化程度和评价精度的有效手段.大量实践证明,机器学习技术能够有效解决测井评价中复杂的非线性问题,目前在测井处理质量和评价精度方面均已取得了一定的突破.但如何更有效利用海量多源测井数据,在繁多的机器学习算法中找到能达到预期结果的最优方法尚未有人进行系统总结.有鉴于此,通过对机器学习算法及其在测井评价中的应用进行详细调研,结合实例系统论述了机器学习在储层测井评价中的分类问题、回归问题、图像处理等方面取得的研究进展,深入探讨了机器学习取得良好应用效果的关键技术,提出了测井精细化储层评价的发展方向.

The core of reservoir logging evaluation is the mathematical modeling of data analysis and model-driven methods. Transforming the logging evaluation process into a machine learning process is an effective means to improve the automation and accuracy of reservoir logging evaluation. A lot of practice has proved that machine learning technology can effectively solve the complex nonlinear problems in logging evaluation. At present, certain breakthroughs have been made in logging processing quality and evaluation accuracy. However, no one has systematically summarized how to use the massive multi-source logging data more effectively and find the best method to achieve the expected results among the various machine learning algorithms. In view of this, this article conducts a detailed investigation on the machine learning algorithm and its application in logging evaluation, and systematically discusses the classification problems, regression problems, and image processing of machine learning in reservoir logging evaluation with examples. Progress, in-depth discussion of the key technologies for achieving good application effects of machine learning, and proposed the development direction of fine logging reservoir evaluation.