气测录井数据处理与油气层判别方法研究

发布时间：2018-05-20 13:15

本文选题：气测录井 + 校正处理　；参考：《东北石油大学》2015年硕士论文

【摘要】：在现代钻井工艺中,工作人员主要通过仪器对钻井产生的各项信息进行采集、分析,由此了解所钻进区域的地质构造情况。随着科技的进步与钻井工艺的不断完善,此项技术逐渐演变成录井技术,到目前为止录井技术已经成为获取此类数据的重要手段。通过检测气测录井数据,工作人员不但可随时监控井下情况,还可以对所钻进的储层流体性质进行判别,判断哪些是油层、哪些是气层等,此项工艺称为气测油气层判别技术。当前采用的气测油气层判别工艺普遍存在两大弊端,其一是气测数据易受很多外在因素影响,所以采用的气测数据都非最初数据,因此降低了对储层判别的准确率。其二是在现代的气测油气层判别方法中,大多都是人工判别方法,在判别结果中人为干扰因素起主导作用,包括气测费歇尔与气测贝叶斯油气层判别方法也存在判别结果无法细化与建模时间较长两种缺陷。因此,本文提出了气测录井数据校正处理方法与基于BP神经网络的气测油气层判别方法。首先,研究影响气测数据准确性的每种因素,将其分为地质环境因素与设备影响因素两大类,并且针对每种影响因素提出相应的气测数据校正处理方法,将仪器所记录的气测数据尽可能恢复到最初状态。其次,采用气测数据与BP神经网络相结合的方法,建立气测BP油气层判别网络模型。针对BP神经网络在建模阶段易出现陷入局部最小值的问题,提出一种改进的自适应学习效率方法应用于网络模型当中,减少了气测BP油气层判别网络模型训练时间,提高了工作效率。对改进的气测BP油气层判别方法进行应用测试。以一组未知判别结果的气测数据作为样本数据,首先用气测数据校正处理方法对其进行校正处理,提高数据的准确性去除环境因素的影响,然后分别采用改进的气测BP油气层判别方法和三角判别方法对其进行判别,根据后期的试油结论与判别结果进行对比,可以看出改进的气测油气层判别方法比三角判别方法具有更高的准确率。
[Abstract]:In modern drilling technology, the workers collect and analyze the information generated by drilling mainly through instruments, and then understand the geological structure of the drilling area. With the progress of science and technology and the continuous improvement of drilling technology, this technology has gradually evolved into a mud logging technology, so far logging technology has become an important means to obtain this kind of data. By detecting the gas logging data, the workers can not only monitor the downhole situation at any time, but also judge the fluid properties of the drilled reservoir, and judge which is the oil layer and which is the gas reservoir. This process is called gas logging oil and gas reservoir discrimination technology. There are two disadvantages in the current gas reservoir identification technology. One is that the gas data are easily affected by many external factors, so the gas data used are not the initial data, so the accuracy of reservoir discrimination is reduced. The other is that most of the modern discriminant methods for gas-bearing oil and gas reservoirs are artificial ones, which play a leading role in discriminating results by human interference factors. There are also two defects in the discriminant method, which include gas measurement Fischer and gas Bayesian oil and gas reservoir discrimination, which can not be refined and the modeling time is longer. Therefore, the method of gas logging data correction and BP neural network is put forward in this paper. First of all, every factor affecting the accuracy of gas data is studied, which is divided into two categories: geological environment factor and equipment influence factor, and the corresponding correction and processing method of gas survey data is put forward for each kind of influence factor. Restore the gas data recorded by the instrument to its original state as far as possible. Secondly, using the method of combining gas data with BP neural network, the model of BP oil and gas reservoir discriminant network is established. In view of the problem that BP neural network is prone to fall into local minimum in the modeling stage, an improved adaptive learning efficiency method is proposed to apply to the network model, which reduces the training time of BP oil and gas layer discriminant network model. Improved working efficiency. The application test of improved BP oil and gas reservoir discrimination method is carried out. Taking a set of gas data of unknown discriminant result as sample data, the method of gas data correction and processing is used to correct the data, and the accuracy of the data is improved to remove the influence of environmental factors. Then the improved BP oil and gas reservoir discrimination method and the triangle discriminant method are used to distinguish them, and the results are compared according to the later oil test results. It can be seen that the improved method has higher accuracy than the triangle method.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE142

【参考文献】