改进人工神经网络原理对储层渗透率的预测——以北部湾盆地涠西南凹陷为例

人工神经网络的计算方法是一种非线性处理系统,是根据测井数据进行储层物性参数预测的方法。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,没有把临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。为弥补这一不足,利用相临多个层位的数据点进行学习,进而建立储层渗透率的预测模型,并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对地层的渗透率进行逐点计算。通过北部湾盆地涠西南凹陷的实例实践表明,用该方法预测的渗透率与实测的渗透率的值符合较好。

Prediction of reservoir permeability with improved artificial neural network principle: taking the Southwest Weizhou Depression in Beibuwan Basin as an example

Computational method of artificial neural network is a non-linear processing system, which predicts reservoir physical property with logging data. In previous calculation of permeability with genetic algorithm, only single data points were used and no data of the neighboring horizons were involved in the training, thus the accuracy and reliability of the prediction model were limited. In order to solve this problem, the data points of several neighboring horizons are used for training and a prediction model of reservoir permeability is built. Based on the normalization of core test data and relevant logging data, reservoir permeability is calculated point-by-point with the improved windowing technique and feedback neural network method. Its application to petroleum exploration in Southwest Weizhou depression of Beibuwan basin shows that the predicted and measured permeability coincides well.