应用人工智能优化降低核磁共振孔隙度测量误差

Improvement of nuclear magnet resource reservoir porosity measuring accuracy by artificial intelligent algorithm

核磁共振测量孔隙度通常偏小,这是因为储层岩石中含有顺磁物质和黏土.为此提出了一种通过采用人工智能算法,根据相关因素对核磁共振测量孔隙度进行校准的新方法.该方法首先根据信息增益的原则,通过数据挖掘找出与核磁孔隙度偏差相关的因素作为神经网络的参数,之后用常规方法测得的孔隙度对神经网络进行训练,并根据实验结果对网络的算法和参数进行优化,最终将实测核磁孔隙度的相对误差从29.35%降低到11.37%.这一结果表明应用人工智能算法能够有效提高核磁共振法测量孔隙度的精度.

The reservoir porosity measured by nuclear magnetic resonance(NMR) is usually less than that measured using regular method,which is because the reservoir rock contains paramagnetic materials and clay.A method decreasing nuclear magnetic resonance porosity measuring error is put forward,which is based on rock's information and using artificial neural network method.Firstly,the factors related to NMR porosity measurement error are found by data mining of a large amount of core sample information based on information gain principle and they are used as the input parameters of artificial network.Then,the artificial neural network is trained with the porosity data of a number of core samples measured by regular method and the algorithm and parameters of the network are optimized according to the porosity data.Finally,the average relative error of the nuclear magnetic resonance porosity measuring is reduced from 29.35% to 11.37% using the optimized network,which shows that the accuracy of NMR porosity measuring can be improved using artificial intelligent algorithm.