基于BP神经网络的储层渗透率预测方法研究

Research on Reservoir Permeability Prediction Method Based on BP Neural Network

传统的间接解释法预测主要采用线性回归方法进行渗透率预测,但此方法存在一个极大的缺点,因为所有数据不一定都是线性关系故采用线性回归方式预测精确度差别较大。针对此问题论文提出基于BP神经网络的储层渗透率预测方法对储层存在的大多数非线性关系数据进行预测。我们首先对选取的原始测井数据进行数据清洗,数据归一化处理;然后,采用BP神经网络算法进行数据特征分析计算,最后使用岩性剖面解释数据验证预测结果。论文采用基于BP神经网络利用测井曲线对储层渗透率预测的方法对松辽盆地构造区某井进行实验,达到了渗透率平均相对预测误差减小、精度大大提高,并且满足测井解释储层参数精度要求。

The traditional indirect interpretation method mainly uses linear regression to predict the parameters,but this meth-od has a great disadvantage,because all the data are not necessarily linear,so the linear regression prediction accuracy varies great-ly.Aiming at this problem,this paper proposes a reservoir permeability prediction method based on BP neural network to predict most of the non-linear relationship data existing in the reservoir.It first cleans the selected original logging data and normalizes the data.Then,the BP neural network algorithm is used to analyze and calculate the data characteristics,and finally the lithological pro-file interpretation data is used to verify the prediction results.In this paper,the method of predicting reservoir permeability based on BP neural network and logging curves is used to conduct experiments on a well in the Songliao Basin structural area.The average rel-ative permeability prediction error is reduced,the accuracy is greatly improved,and the logging interpretation is satisfied.Layer pa-rameter accuracy requirements.