多核核函数方法在储层岩性识别中的应用

发布时间：2018-12-25 15:46

【摘要】：我国现有的主力油田,基本都已经进入了中后期的开发阶段,剩余可采油气藏往往是地层状况复杂的非常规油气藏,开采难度非常大。准确识别储层岩性对储层评价有着重要的意义,能够加快勘探开发的进程,促进油气稳产工作。本课题利用采油厂累积的丰富的地质资料数据,结合人工智能技术,提出新的储层岩性识别模型,为油气田的勘探与开发提供有力的支持。本文分析了目前储层岩性识别常用技术的方法与原理。其中最准确的岩性识别方法是通过取心对岩心薄片进行鉴定,但是取心成本非常高,耗时耗力,对每口井都进行取心是不现实的,因此不能满足实际工作需求。后来发展到聚类分析法与主成分分析法等,这些方法解决了不能取心的问题,也取得了较好的结果,但是岩性识别的精度还是不够,针对这些问题,本文提出基于人工智能技术的智能识别方法。本文主要采用测井曲线来识别岩性,常规的智能模型输入输出都是几何点式的,没有考虑到测井曲线随深度变化而变化,因此需要采用能够直接处理过程信号的模型。实际工区部分储层非均质性严重,导致岩性识别正确率还不够高,本文提出多核核函数方法,用多个核函数更加精确的描述储层岩性特征以提高识别率。本文将上述理论相结合提出了多核过程支持向量机与多尺度核径向基过程神经网络两种智能模型来做岩性识别工作。多核模型从机制上改善模型对复杂特征信号的表示能力,并运用智能算法对模型参数进行优化,从而达到更高的识别精度。本课题基于人工智能技术,将理论研究构建成可实际应用的软件原型系统,并用实际数据在系统中进行了应用,且取得了良好的效果。本文研究成果对于油气田的勘探开发工作是有实际使用意义的,能够在一定程度上对实际工作起到帮助的作用,具有理论价值与实际应用价值。
[Abstract]:The main oil fields in our country have basically entered the middle and late stage of development, and the remaining recoverable reservoirs are often unconventional reservoirs with complex stratigraphic conditions, which is very difficult to exploit. Accurate identification of reservoir lithology is of great significance to reservoir evaluation, which can speed up the process of exploration and development and promote the stable production of oil and gas. Based on the abundant geological data accumulated in oil production plant and artificial intelligence technology, a new reservoir lithology identification model is proposed in this paper, which can provide strong support for exploration and development of oil and gas fields. In this paper, the methods and principles of common techniques for reservoir lithology identification are analyzed. The most accurate lithologic identification method is to identify the core slice by coring, but the cost of coring is very high, and it is time-consuming and labor-consuming, so it is not realistic to coring every well, so it can not meet the practical requirements. Later, cluster analysis and principal component analysis were developed. These methods solved the problem of uncoring and achieved good results, but the accuracy of lithology recognition was not enough. This paper presents an intelligent recognition method based on artificial intelligence technology. In this paper, logging curves are mainly used to identify lithology. The conventional intelligent model is geometric point type and does not take into account the change of logging curve with depth. Therefore, it is necessary to adopt a model that can process signals directly. Due to the serious heterogeneity of some reservoirs in practical working areas, the correct rate of lithology recognition is not high enough. In this paper, a multi-kernel function method is proposed to describe the lithologic characteristics of reservoirs more accurately in order to improve the recognition rate. In this paper, two intelligent models of multi-kernel process support vector machine and multi-scale radial basis function neural network are proposed to identify lithology. The multi-core model can improve the representation ability of the model to the complex feature signal from the mechanism, and use the intelligent algorithm to optimize the model parameters, so as to achieve higher recognition accuracy. Based on artificial intelligence technology, this paper constructs a practical software prototype system based on artificial intelligence technology, and uses the actual data in the system, and obtains good results. The research results in this paper are of practical significance to the exploration and development of oil and gas fields, and can help the actual work to a certain extent, and have theoretical value and practical application value.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TE311

【相似文献】