要建成一个匹配油田生产数据的油藏模拟模型,耗费时间极多,已成为一个行业性的挑战。为了辅助历史拟合,目前已实现的有两个工作流程,一个工作流程是在全局优化的条件下通过最小化数模和历史数据之间的错配率来实现,而另一个工作流程是通过训练神经网络产生理想的非线性替代模型。这是一个来源于实验设计的小的数值模拟,可用来减少模拟的次数。 Building a reservoir simulation model that matches production data from oilfields has taken a significant amount of time and has become an industry challenge. In order to assist history fitting, there are two work processes that have been implemented so far. One work flow is achieved under the condition of global optimization by minimizing the mismatch rate between the digital model and the historical data, and the other work flow is through Training neural networks produces an ideal non-linear alternative model. This is a small numerical simulation derived from the experimental design that can be used to reduce the number of simulations.