摘要
由于储层的非均匀性,传统的方法很难得到真实反映储层特性的结果。采用了遗传算法与BP神经网络相结合,利用遗传算法的全局寻优特点,优化神经网络的连接权值和阈值,从而提高网络的训练精度和预测精度。将相似度的概念引入到测井中,定义了相似度在测井中的计算公式,提出了相似度与遗传神经网络相结合的方法。根据取心井段储层物性与测井信息的关系,选取相应的测井曲线,运用MATLAB中神经网络工具箱建立神经网络模型并训练。实例研究表明,预测准确性较高,且可以有效地控制预测精度,避免了因储层差别大而造成的预测精度降低的现象。
Due to the anisotropy of reservoir,using the linear method is difficult to get truly result of the reservoir characteristics.Genetic algorithm and BP neural network are combined,and used the global random hunting function of the genetic algorithm to optimize neural network connection weights and threshold.Threshold.At the same time,the similarity in well logging and calculation formula is defined,and combined similarity and Genetic Neural Network.Considering the relationship between reservoir the physica1 properties and well logging data,the constructed BP neural network model is iterated by choosing corresponding well logging data and using functions of neural network toolbox offered by MATLAB.Compared with real samples,the predictive accuracy is high,and through this method,precision can be effectively control,and avoid the phenomenon that neural network prediction accuracy reduce.
出处
《科学技术与工程》
2011年第35期8846-8850,共5页
Science Technology and Engineering
