论文部分内容阅读
渗透率是油藏描述和油藏工程中较为关键性的参数,因而如何求取较为精确的地层渗透率参数值显得格外重要。本文在岩心分析化验数据和相关测井曲线数据归一化的基础上,利用改进的开窗技术,借助反馈的神经网络方法逐点计算地层的渗透率。以往在利用遗传算法预测渗透率的时候,因为只考虑了单一的数据点,没有把邻近层位的数据加入学习过程中来,故影响了预测模型的精度和可信度。笔者经过系统的研究,用相邻5个层位的数据点进行学习,建立储层渗透率的预测模型。大庆萨尔图油田葡萄花油层组PⅠ1—PⅠ4小层砂岩的油气勘探实践证明,预测的渗透率与实测的渗透率的值符合较好。
Permeability is the more critical parameter in reservoir description and reservoir engineering. Therefore, it is particularly important to find out the more accurate parameters of formation permeability. In this paper, based on the normalization of the core analysis laboratory data and the related well logging data, the improved windowing technology is used to calculate the permeability of the formation point by point by means of feedback neural network. In the past when using genetic algorithm to predict permeability, because only a single data point was considered and no data from adjacent layers was added to the learning process, the accuracy and credibility of the prediction model were affected. After systematic research, the author uses data points of five adjacent layers to study and establishes a prediction model of reservoir permeability. The oil and gas exploration practice of PⅠ1-PⅠ4 small layer sandstone in Putaohua oil-bearing formation of Saertu Oilfield in Daqing shows that the predicted permeability is in good agreement with the measured permeability.