基于BP神经网络原理的人造岩心配比设计模型

Artificial Core Matching Design Model Based on BP Neural Network

利用人造岩心模拟已知物性参数的地层条件来替代天然岩心,已成为天然岩心存量较少条件下石油开发室内研究的一种趋势。通过综合全面地考虑人造岩心物性参数的影响因素,制定人造岩心配比设计正交方案,制备人造岩心,测试物性参数,分析预考虑影响因素与物性参数的关系;采用灰色关联法,明确岩心孔隙度、渗透率、粒度中值的主要控制参数,进一步分析和判断预考虑影响因素的合理性;基于影响因素分析结果、实验数据、BP神经网络原理,建立人造岩心配比设计数学模型。结果表明,预考虑影响因素的全面性和影响因素数据化、定量化是建立模型的基础。影响人造岩心物性参数的影响因素主要有砂型配比、胶结物加量、压制压力和加压时间等。其中,胶结剂加量和压制压力对孔隙度影响程度大;粒径为0.224数0.45 mm石英砂加量对渗透率影响最大;0.154数0.28 mm石英砂加量对粒度中值影响最大;0.074数0.18 mm石英砂加量对孔隙度、渗透率和粒度中值的影响最弱。由配比设计模型计算岩心制备加量,根据计算结果制备的人造岩心物性参数测试值与期望值的总体相对误差小于10%。该方法可用于指导定制物性参数模拟误差小的人造岩心。

Artificial cores were used to simulate the formation conditions of known physical parameters and replaced natural cores, which had become a trend of oil development laboratory research under the condition of less natural cores stock. Based on the comprehensive consideration of the factors affecting the physical parameters of artificial cores,the orthogonal scheme for the proportion design of artificial cores was formulated. This scheme could be used as the instruction of making artificial cores. By testing the physical parameters of cores,the relationship between the influence factors and the physical parameters was analyzed. Using grey correlation method to determine the main control parameter of porosity,permeability and median size,and to further analyze and judge the reasonableness of pre-consideration factors. Based on the results of influencing factors analysis,experimental data and the principle of BP neural network,a mathematical model of artificial core matching design was established. The results showed that the comprehensiveness of pre-consideration factors,the digitization and quantification of influencing factors were the foundation of the model. The main factors affecting the physical parameters of artificial cores were sand ratio,the amount of cement,pressing pressure and time,etc. And the amount of cement and pressing pressure had great influence on porosity. The amount of quartz sand with 0.224-0.45 mm particle size had the biggest effect on the permeability,and that with 0.154-0.28 mm particle size had the greatest influence on the median size. The effect of quartz sand dosage with 0.074-0.18 mm particle size on porosity,permeability and median size was the weakest. This model was used to calculate the amount of core preparation,and the average relative error between measured values and expected values of the physical parameters of artificial cores was less than 10%. This method could be used as the instruction of making artificial cores with small simulation error of physical parameters.