神经网络在油藏数值模拟中的应用

Application of Neural Network in Reservoir Simulation

油藏数值模拟的主要内容是从产量方面预测经济效益,目标是用最小的投资获得最便捷的开采方式进而从油藏中得到最合适的油气资源配置,油藏实际开采中的非均质性导致水淹越来越严重。因此在再现油田的实际开采过程时数据更加复杂,所以神经网络在油藏数值模拟中得到广泛应用。尤其在解释渗透率、饱和度、孔隙度等参数和分析油层水淹情况、岩性油藏的含油边界等方面的应用更为普遍。与常规方法相比,不需要建立计算公式和具体模型,有很强的预测精度和适应性。

The main content of oil reservoir numerical simulation is to forecast economic benefits from production aspects, the target is to obtain the most convenient development way with minimal investment. But in actual reservoir due to the heterogeneity of formation, water out is more and more serious. Thus reappearing actual production process data of oil field is much more complicated, so the neural network is widely used in reservoir numerical simulation, especially in explanation of parameters such as saturation, porosity and permeability, analysis of water flooding and oil boundary of lithologic reservoir. Compared with conventional methods, calculation formula and specific model need not be established in using neural networks, and it has high prediction precision and adaptability.